Unnamed: 0
int64 0
5k
| title
stringlengths 9
210
| abstract
stringlengths 164
1.92k
|
---|---|---|
4,500 | Interpolating Compressed Parameter Subspaces | Inspired by recent work on neural subspaces and mode connectivity, we revisit
parameter subspace sampling for shifted and/or interpolatable input
distributions (instead of a single, unshifted distribution). We enforce a
compressed geometric structure upon a set of trained parameters mapped to a set
of train-time distributions, denoting the resulting subspaces as Compressed
Parameter Subspaces (CPS). We show the success and failure modes of the types
of shifted distributions whose optimal parameters reside in the CPS. We find
that ensembling point-estimates within a CPS can yield a high average accuracy
across a range of test-time distributions, including backdoor, adversarial,
permutation, stylization and rotation perturbations. We also find that the CPS
can contain low-loss point-estimates for various task shifts (albeit
interpolated, perturbed, unseen or non-identical coarse labels). We further
demonstrate this property in a continual learning setting with CIFAR100. |
4,501 | Time Series Anomaly Detection via Reinforcement Learning-Based Model Selection | Time series anomaly detection is of critical importance for the reliable and
efficient operation of real-world systems. Many anomaly detection models have
been developed throughout the years based on various assumptions regarding
anomaly characteristics. However, due to the complex nature of real-world data,
different anomalies within a time series usually have diverse profiles
supporting different anomaly assumptions, making it difficult to find a single
anomaly detector that can consistently beat all other models. In this work, to
harness the benefits of different base models, we assume that a pool of anomaly
detection models is accessible and propose to utilize reinforcement learning to
dynamically select a candidate model from these base models. Experiments on
real-world data have been implemented. It is demonstrated that the proposed
strategy can outperforms all baseline models in terms of overall performance. |
4,502 | A Rule Search Framework for the Early Identification of Chronic Emergency Homeless Shelter Clients | This paper uses rule search techniques for the early identification of
emergency homeless shelter clients who are at risk of becoming long term or
chronic shelter users. Using a data set from a major North American shelter
containing 12 years of service interactions with over 40,000 individuals, the
optimized pruning for unordered search (OPUS) algorithm is used to develop
rules that are both intuitive and effective. The rules are evaluated within a
framework compatible with the real-time delivery of a housing program meant to
transition high risk clients to supportive housing. Results demonstrate that
the median time to identification of clients at risk of chronic shelter use
drops from 297 days to 162 days when the methods in this paper are applied. |
4,503 | Beyond Labels: Visual Representations for Bone Marrow Cell Morphology Recognition | Analyzing and inspecting bone marrow cell cytomorphology is a critical but
highly complex and time-consuming component of hematopathology diagnosis.
Recent advancements in artificial intelligence have paved the way for the
application of deep learning algorithms to complex medical tasks. Nevertheless,
there are many challenges in applying effective learning algorithms to medical
image analysis, such as the lack of sufficient and reliably annotated training
datasets and the highly class-imbalanced nature of most medical data. Here, we
improve on the state-of-the-art methodologies of bone marrow cell recognition
by deviating from sole reliance on labeled data and leveraging self-supervision
in training our learning models. We investigate our approach's effectiveness in
identifying bone marrow cell types. Our experiments demonstrate significant
performance improvements in conducting different bone marrow cell recognition
tasks compared to the current state-of-the-art methodologies. |
4,504 | Real Time Multi-Object Detection for Helmet Safety | The National Football League and Amazon Web Services teamed up to develop the
best sports injury surveillance and mitigation program via the Kaggle
competition. Through which the NFL wants to assign specific players to each
helmet, which would help accurately identify each player's "exposures"
throughout a football play. We are trying to implement a computer vision based
ML algorithms capable of assigning detected helmet impacts to correct players
via tracking information. Our paper will explain the approach to automatically
track player helmets and their collisions. This will also allow them to review
previous plays and explore the trends in exposure over time. |
4,505 | Incremental Learning with Differentiable Architecture and Forgetting Search | As progress is made on training machine learning models on incrementally
expanding classification tasks (i.e., incremental learning), a next step is to
translate this progress to industry expectations. One technique missing from
incremental learning is automatic architecture design via Neural Architecture
Search (NAS). In this paper, we show that leveraging NAS for incremental
learning results in strong performance gains for classification tasks.
Specifically, we contribute the following: first, we create a strong baseline
approach for incremental learning based on Differentiable Architecture Search
(DARTS) and state-of-the-art incremental learning strategies, outperforming
many existing strategies trained with similar-sized popular architectures;
second, we extend the idea of architecture search to regularize architecture
forgetting, boosting performance past our proposed baseline. We evaluate our
method on both RF signal and image classification tasks, and demonstrate we can
achieve up to a 10% performance increase over state-of-the-art methods. Most
importantly, our contribution enables learning from continuous distributions on
real-world application data for which the complexity of the data distribution
is unknown, or the modality less explored (such as RF signal classification). |
4,506 | Content-Context Factorized Representations for Automated Speech Recognition | Deep neural networks have largely demonstrated their ability to perform
automated speech recognition (ASR) by extracting meaningful features from input
audio frames. Such features, however, may consist not only of information about
the spoken language content, but also may contain information about unnecessary
contexts such as background noise and sounds or speaker identity, accent, or
protected attributes. Such information can directly harm generalization
performance, by introducing spurious correlations between the spoken words and
the context in which such words were spoken. In this work, we introduce an
unsupervised, encoder-agnostic method for factoring speech-encoder
representations into explicit content-encoding representations and spurious
context-encoding representations. By doing so, we demonstrate improved
performance on standard ASR benchmarks, as well as improved performance in both
real-world and artificially noisy ASR scenarios. |
4,507 | Transformer with Memory Replay | Transformers achieve state-of-the-art performance for natural language
processing tasks by pre-training on large-scale text corpora. They are
extremely compute-intensive and have very high sample complexity. Memory replay
is a mechanism that remembers and reuses past examples by saving to and
replaying from a memory buffer. It has been successfully used in reinforcement
learning and GANs due to better sample efficiency. In this paper, we propose
\emph{Transformer with Memory Replay} (TMR), which integrates memory replay
with transformer, making transformer more sample-efficient. Experiments on GLUE
and SQuAD benchmark datasets show that Transformer with Memory Replay achieves
at least $1\%$ point increase compared to the baseline transformer model when
pretrained with the same number of examples. Further, by adopting a careful
design that reduces the wall-clock time overhead of memory replay, we also
empirically achieve a better runtime efficiency. |
4,508 | Service Delay Minimization for Federated Learning over Mobile Devices | Federated learning (FL) over mobile devices has fostered numerous intriguing
applications/services, many of which are delay-sensitive. In this paper, we
propose a service delay efficient FL (SDEFL) scheme over mobile devices. Unlike
traditional communication efficient FL, which regards wireless communications
as the bottleneck, we find that under many situations, the local computing
delay is comparable to the communication delay during the FL training process,
given the development of high-speed wireless transmission techniques. Thus, the
service delay in FL should be computing delay + communication delay over
training rounds. To minimize the service delay of FL, simply reducing local
computing/communication delay independently is not enough. The delay trade-off
between local computing and wireless communications must be considered.
Besides, we empirically study the impacts of local computing control and
compression strategies (i.e., the number of local updates, weight quantization,
and gradient quantization) on computing, communication and service delays.
Based on those trade-off observation and empirical studies, we develop an
optimization scheme to minimize the service delay of FL over heterogeneous
devices. We establish testbeds and conduct extensive emulations/experiments to
verify our theoretical analysis. The results show that SDEFL reduces notable
service delay with a small accuracy drop compared to peer designs. |
4,509 | Automated Scoring for Reading Comprehension via In-context BERT Tuning | Automated scoring of open-ended student responses has the potential to
significantly reduce human grader effort. Recent advances in automated scoring
often leverage textual representations based on pre-trained language models
such as BERT and GPT as input to scoring models. Most existing approaches train
a separate model for each item/question, which is suitable for scenarios such
as essay scoring where items can be quite different from one another. However,
these approaches have two limitations: 1) they fail to leverage item linkage
for scenarios such as reading comprehension where multiple items may share a
reading passage; 2) they are not scalable since storing one model per item
becomes difficult when models have a large number of parameters. In this paper,
we report our (grand prize-winning) solution to the National Assessment of
Education Progress (NAEP) automated scoring challenge for reading
comprehension. Our approach, in-context BERT fine-tuning, produces a single
shared scoring model for all items with a carefully-designed input structure to
provide contextual information on each item. We demonstrate the effectiveness
of our approach via local evaluations using the training dataset provided by
the challenge. We also discuss the biases, common error types, and limitations
of our approach. |
4,510 | Recurrent segmentation meets block models in temporal networks | A popular approach to model interactions is to represent them as a network
with nodes being the agents and the interactions being the edges. Interactions
are often timestamped, which leads to having timestamped edges. Many real-world
temporal networks have a recurrent or possibly cyclic behaviour. For example,
social network activity may be heightened during certain hours of day. In this
paper, our main interest is to model recurrent activity in such temporal
networks. As a starting point we use stochastic block model, a popular choice
for modelling static networks, where nodes are split into $R$ groups. We extend
this model to temporal networks by modelling the edges with a Poisson process.
We make the parameters of the process dependent on time by segmenting the time
line into $K$ segments. To enforce the recurring activity we require that only
$H < K$ different set of parameters can be used, that is, several, not
necessarily consecutive, segments must share their parameters. We prove that
the searching for optimal blocks and segmentation is an NP-hard problem.
Consequently, we split the problem into 3 subproblems where we optimize blocks,
model parameters, and segmentation in turn while keeping the remaining
structures fixed. We propose an iterative algorithm that requires $O(KHm + Rn +
R^2H)$ time per iteration, where $n$ and $m$ are the number of nodes and edges
in the network. We demonstrate experimentally that the number of required
iterations is typically low, the algorithm is able to discover the ground truth
from synthetic datasets, and show that certain real-world networks exhibit
recurrent behaviour as the likelihood does not deteriorate when $H$ is lowered. |
4,511 | Mean-Field Analysis of Two-Layer Neural Networks: Global Optimality with Linear Convergence Rates | We consider optimizing two-layer neural networks in the mean-field regime
where the learning dynamics of network weights can be approximated by the
evolution in the space of probability measures over the weight parameters
associated with the neurons. The mean-field regime is a theoretically
attractive alternative to the NTK (lazy training) regime which is only
restricted locally in the so-called neural tangent kernel space around
specialized initializations. Several prior works (\cite{mei2018mean,
chizat2018global}) establish the asymptotic global optimality of the mean-field
regime, but it is still challenging to obtain a quantitative convergence rate
due to the complicated nonlinearity of the training dynamics. This work
establishes a new linear convergence result for two-layer neural networks
trained by continuous-time noisy gradient descent in the mean-field regime. Our
result relies on a novelty logarithmic Sobolev inequality for two-layer neural
networks, and uniform upper bounds on the logarithmic Sobolev constants for a
family of measures determined by the evolving distribution of hidden neurons. |
4,512 | MCVD: Masked Conditional Video Diffusion for Prediction, Generation, and Interpolation | Video prediction is a challenging task. The quality of video frames from
current state-of-the-art (SOTA) generative models tends to be poor and
generalization beyond the training data is difficult. Furthermore, existing
prediction frameworks are typically not capable of simultaneously handling
other video-related tasks such as unconditional generation or interpolation. In
this work, we devise a general-purpose framework called Masked Conditional
Video Diffusion (MCVD) for all of these video synthesis tasks using a
probabilistic conditional score-based denoising diffusion model, conditioned on
past and/or future frames. We train the model in a manner where we randomly and
independently mask all the past frames or all the future frames. This novel but
straightforward setup allows us to train a single model that is capable of
executing a broad range of video tasks, specifically: future/past prediction --
when only future/past frames are masked; unconditional generation -- when both
past and future frames are masked; and interpolation -- when neither past nor
future frames are masked. Our experiments show that this approach can generate
high-quality frames for diverse types of videos. Our MCVD models are built from
simple non-recurrent 2D-convolutional architectures, conditioning on blocks of
frames and generating blocks of frames. We generate videos of arbitrary lengths
autoregressively in a block-wise manner. Our approach yields SOTA results
across standard video prediction and interpolation benchmarks, with computation
times for training models measured in 1-12 days using $\le$ 4 GPUs. Project
page: https://mask-cond-video-diffusion.github.io ; Code :
https://github.com/voletiv/mcvd-pytorch |
4,513 | Deconfounding Actor-Critic Network with Policy Adaptation for Dynamic Treatment Regimes | Despite intense efforts in basic and clinical research, an individualized
ventilation strategy for critically ill patients remains a major challenge.
Recently, dynamic treatment regime (DTR) with reinforcement learning (RL) on
electronic health records (EHR) has attracted interest from both the healthcare
industry and machine learning research community. However, most learned DTR
policies might be biased due to the existence of confounders. Although some
treatment actions non-survivors received may be helpful, if confounders cause
the mortality, the training of RL models guided by long-term outcomes (e.g.,
90-day mortality) would punish those treatment actions causing the learned DTR
policies to be suboptimal. In this study, we develop a new deconfounding
actor-critic network (DAC) to learn optimal DTR policies for patients. To
alleviate confounding issues, we incorporate a patient resampling module and a
confounding balance module into our actor-critic framework. To avoid punishing
the effective treatment actions non-survivors received, we design a short-term
reward to capture patients' immediate health state changes. Combining
short-term with long-term rewards could further improve the model performance.
Moreover, we introduce a policy adaptation method to successfully transfer the
learned model to new-source small-scale datasets. The experimental results on
one semi-synthetic and two different real-world datasets show the proposed
model outperforms the state-of-the-art models. The proposed model provides
individualized treatment decisions for mechanical ventilation that could
improve patient outcomes. |
4,514 | Confident Clustering via PCA Compression Ratio and Its Application to Single-cell RNA-seq Analysis | Unsupervised clustering algorithms for vectors has been widely used in the
area of machine learning. Many applications, including the biological data we
studied in this paper, contain some boundary datapoints which show combination
properties of two underlying clusters and could lower the performance of the
traditional clustering algorithms. We develop a confident clustering method
aiming to diminish the influence of these datapoints and improve the clustering
results. Concretely, for a list of datapoints, we give two clustering results.
The first-round clustering attempts to classify only pure vectors with high
confidence. Based on it, we classify more vectors with less confidence in the
second round. We validate our algorithm on single-cell RNA-seq data, which is a
powerful and widely used tool in biology area. Our confident clustering shows a
high accuracy on our tested datasets. In addition, unlike traditional
clustering methods in single-cell analysis, the confident clustering shows high
stability under different choices of parameters. |
4,515 | A Hardware-Aware Framework for Accelerating Neural Architecture Search Across Modalities | Recent advances in Neural Architecture Search (NAS) such as one-shot NAS
offer the ability to extract specialized hardware-aware sub-network
configurations from a task-specific super-network. While considerable effort
has been employed towards improving the first stage, namely, the training of
the super-network, the search for derivative high-performing sub-networks is
still under-explored. Popular methods decouple the super-network training from
the sub-network search and use performance predictors to reduce the
computational burden of searching on different hardware platforms. We propose a
flexible search framework that automatically and efficiently finds optimal
sub-networks that are optimized for different performance metrics and hardware
configurations. Specifically, we show how evolutionary algorithms can be paired
with lightly trained objective predictors in an iterative cycle to accelerate
architecture search in a multi-objective setting for various modalities
including machine translation and image classification. |
4,516 | Preliminary study on the impact of EEG density on TMS-EEG classification in Alzheimer's disease | Transcranial magnetic stimulation co-registered with electroencephalographic
(TMS-EEG) has previously proven a helpful tool in the study of Alzheimer's
disease (AD). In this work, we investigate the use of TMS-evoked EEG responses
to classify AD patients from healthy controls (HC). By using a dataset
containing 17AD and 17HC, we extract various time domain features from
individual TMS responses and average them over a low, medium and high density
EEG electrode set. Within a leave-one-subject-out validation scenario, the best
classification performance for AD vs. HC was obtained using a high-density
electrode with a Random Forest classifier. The accuracy, sensitivity and
specificity were of 92.7%, 96.58% and 88.2% respectively. |
4,517 | A toolbox for idea generation and evaluation: Machine learning, data-driven, and contest-driven approaches to support idea generation | The significance and abundance of data are increasing due to the growing
digital data generated from social media, sensors, scholarly literature,
patents, different forms of documents published online, databases, product
manuals, etc. Various data sources can be used to generate ideas, yet, in
addition to bias, the size of the available digital data is a major challenge
when it comes to manual analysis. Hence, human-machine interaction is essential
for generating valuable ideas where machine learning and data-driven techniques
generate patterns from data and serve human sense-making. However, the use of
machine learning and data-driven approaches to generate ideas is a relatively
new area. Moreover, it is also possible to stimulate innovation using
contest-driven idea generation and evaluation. The results and contributions of
this thesis can be viewed as a toolbox of idea-generation techniques, including
a list of data-driven and machine learning techniques with corresponding data
sources and models to support idea generation. In addition, the results include
two models, one method and one framework, to better support data-driven and
contest- driven idea generation. The beneficiaries of these artefacts are
practitioners in data and knowledge engineering, data mining project managers,
and innovation agents. Innovation agents include incubators, contest
organizers, consultants, innovation accelerators, and industries. Since the
proposed artefacts consist of process models augmented with AI techniques,
human-centred AI is a promising area of research that can contribute to the
artefacts' further development and promote creativity. |
4,518 | HyBNN and FedHyBNN: (Federated) Hybrid Binary Neural Networks | Binary Neural Networks (BNNs), neural networks with weights and activations
constrained to -1(0) and +1, are an alternative to deep neural networks which
offer faster training, lower memory consumption and lightweight models, ideal
for use in resource constrained devices while being able to utilize the
architecture of their deep neural network counterpart. However, the input
binarization step used in BNNs causes a severe accuracy loss. In this paper, we
introduce a novel hybrid neural network architecture, Hybrid Binary Neural
Network (HyBNN), consisting of a task-independent, general, full-precision
variational autoencoder with a binary latent space and a task specific binary
neural network that is able to greatly limit the accuracy loss due to input
binarization by using the full precision variational autoencoder as a feature
extractor. We use it to combine the state-of-the-art accuracy of deep neural
networks with the much faster training time, quicker test-time inference and
power efficiency of binary neural networks. We show that our proposed system is
able to very significantly outperform a vanilla binary neural network with
input binarization. We also introduce FedHyBNN, a highly communication
efficient federated counterpart to HyBNN and demonstrate that it is able to
reach the same accuracy as its non-federated equivalent. We make our source
code, experimental parameters and models available at:
https://anonymous.4open.science/r/HyBNN. |
4,519 | Why GANs are overkill for NLP | This work offers a novel theoretical perspective on why, despite numerous
attempts, adversarial approaches to generative modeling (e.g., GANs) have not
been as popular for certain generation tasks, particularly sequential tasks
such as Natural Language Generation, as they have in others, such as Computer
Vision. In particular, on sequential data such as text, maximum-likelihood
approaches are significantly more utilized than GANs. We show that, while it
may seem that maximizing likelihood is inherently different than minimizing
distinguishability, this distinction is largely artificial and only holds for
limited models. We argue that minimizing KL-divergence (i.e., maximizing
likelihood) is a more efficient approach to effectively minimizing the same
distinguishability criteria that adversarial models seek to optimize.
Reductions show that minimizing distinguishability can be seen as simply
boosting likelihood for certain families of models including n-gram models and
neural networks with a softmax output layer. To achieve a full polynomial-time
reduction, a novel next-token distinguishability model is considered. |
4,520 | Summarization as Indirect Supervision for Relation Extraction | Relation extraction (RE) models have been challenged by their reliance on
training data with expensive annotations. Considering that summarization tasks
aim at acquiring concise expressions of synoptical information from the longer
context, these tasks naturally align with the objective of RE, i.e., extracting
a kind of synoptical information that describes the relation of entity
mentions. We present SuRE, which converts RE into a summarization formulation.
SuRE leads to more precise and resource-efficient RE based on indirect
supervision from summarization tasks. To achieve this goal, we develop sentence
and relation conversion techniques that essentially bridge the formulation of
summarization and RE tasks. We also incorporate constraint decoding techniques
with Trie scoring to further enhance summarization-based RE with robust
inference. Experiments on three RE datasets demonstrate the effectiveness of
SuRE in both full-dataset and low-resource settings, showing that summarization
is a promising source of indirect supervision to improve RE models. |
4,521 | Concurrent Policy Blending and System Identification for Generalized Assistive Control | In this work, we address the problem of solving complex collaborative robotic
tasks subject to multiple varying parameters. Our approach combines
simultaneous policy blending with system identification to create generalized
policies that are robust to changes in system parameters. We employ a blending
network whose state space relies solely on parameter estimates from a system
identification technique. As a result, this blending network learns how to
handle parameter changes instead of trying to learn how to solve the task for a
generalized parameter set simultaneously. We demonstrate our scheme's ability
on a collaborative robot and human itching task in which the human has motor
impairments. We then showcase our approach's efficiency with a variety of
system identification techniques when compared to standard domain
randomization. |
4,522 | Classification of Intra-Pulse Modulation of Radar Signals by Feature Fusion Based Convolutional Neural Networks | Detection and classification of radars based on pulses they transmit is an
important application in electronic warfare systems. In this work, we propose a
novel deep-learning based technique that automatically recognizes intra-pulse
modulation types of radar signals. Re-assigned spectrogram of measured radar
signal and detected outliers of its instantaneous phases filtered by a special
function are used for training multiple convolutional neural networks.
Automatically extracted features from the networks are fused to distinguish
frequency and phase modulated signals. Simulation results show that the
proposed FF-CNN (Feature Fusion based Convolutional Neural Network) technique
outperforms the current state-of-the-art alternatives and is easily scalable
among broad range of modulation types. |
4,523 | Learning Interface Conditions in Domain Decomposition Solvers | Domain decomposition methods are widely used and effective in the
approximation of solutions to partial differential equations. Yet the optimal
construction of these methods requires tedious analysis and is often available
only in simplified, structured-grid settings, limiting their use for more
complex problems. In this work, we generalize optimized Schwarz domain
decomposition methods to unstructured-grid problems, using Graph Convolutional
Neural Networks (GCNNs) and unsupervised learning to learn optimal
modifications at subdomain interfaces. A key ingredient in our approach is an
improved loss function, enabling effective training on relatively small
problems, but robust performance on arbitrarily large problems, with
computational cost linear in problem size. The performance of the learned
linear solvers is compared with both classical and optimized domain
decomposition algorithms, for both structured- and unstructured-grid problems. |
4,524 | Capturing cross-session neural population variability through self-supervised identification of consistent neuron ensembles | Decoding stimuli or behaviour from recorded neural activity is a common
approach to interrogate brain function in research, and an essential part of
brain-computer and brain-machine interfaces. Reliable decoding even from small
neural populations is possible because high dimensional neural population
activity typically occupies low dimensional manifolds that are discoverable
with suitable latent variable models. Over time however, drifts in activity of
individual neurons and instabilities in neural recording devices can be
substantial, making stable decoding over days and weeks impractical. While this
drift cannot be predicted on an individual neuron level, population level
variations over consecutive recording sessions such as differing sets of
neurons and varying permutations of consistent neurons in recorded data may be
learnable when the underlying manifold is stable over time. Classification of
consistent versus unfamiliar neurons across sessions and accounting for
deviations in the order of consistent recording neurons in recording datasets
over sessions of recordings may then maintain decoding performance. In this
work we show that self-supervised training of a deep neural network can be used
to compensate for this inter-session variability. As a result, a sequential
autoencoding model can maintain state-of-the-art behaviour decoding performance
for completely unseen recording sessions several days into the future. Our
approach only requires a single recording session for training the model, and
is a step towards reliable, recalibration-free brain computer interfaces. |
4,525 | Algorithms for Weak Optimal Transport with an Application to Economics | The theory of weak optimal transport (WOT), introduced by [Gozlan et al.,
2017], generalizes the classic Monge-Kantorovich framework by allowing the
transport cost between one point and the points it is matched with to be
nonlinear. In the so-called barycentric version of WOT, the cost for
transporting a point $x$ only depends on $x$ and on the barycenter of the
points it is matched with. This aggregation property of WOT is appealing in
machine learning, economics and finance. Yet algorithms to compute WOT have
only been developed for the special case of quadratic barycentric WOT, or
depend on neural networks with no guarantee on the computed value and matching.
The main difficulty lies in the transportation constraints which are costly to
project onto. In this paper, we propose to use mirror descent algorithms to
solve the primal and dual versions of the WOT problem. We also apply our
algorithms to the variant of WOT introduced by [Chon\'e et al., 2022] where
mass is distributed from one space to another through unnormalized kernels
(WOTUK). We empirically compare the solutions of WOT and WOTUK with classical
OT. We illustrate our numerical methods to the economic framework of [Chon\'e
and Kramarz, 2021], namely the matching between workers and firms on labor
markets. |
4,526 | Deep Learning Methods for Proximal Inference via Maximum Moment Restriction | The No Unmeasured Confounding Assumption is widely used to identify causal
effects in observational studies. Recent work on proximal inference has
provided alternative identification results that succeed even in the presence
of unobserved confounders, provided that one has measured a sufficiently rich
set of proxy variables, satisfying specific structural conditions. However,
proximal inference requires solving an ill-posed integral equation. Previous
approaches have used a variety of machine learning techniques to estimate a
solution to this integral equation, commonly referred to as the bridge
function. However, prior work has often been limited by relying on
pre-specified kernel functions, which are not data adaptive and struggle to
scale to large datasets. In this work, we introduce a flexible and scalable
method based on a deep neural network to estimate causal effects in the
presence of unmeasured confounding using proximal inference. Our method
achieves state of the art performance on two well-established proximal
inference benchmarks. Finally, we provide theoretical consistency guarantees
for our method. |
4,527 | A Learning-Based Approach to Approximate Coded Computation | Lagrange coded computation (LCC) is essential to solving problems about
matrix polynomials in a coded distributed fashion; nevertheless, it can only
solve the problems that are representable as matrix polynomials. In this paper,
we propose AICC, an AI-aided learning approach that is inspired by LCC but also
uses deep neural networks (DNNs). It is appropriate for coded computation of
more general functions. Numerical simulations demonstrate the suitability of
the proposed approach for the coded computation of different matrix functions
that are often utilized in digital signal processing. |
4,528 | MiDAS: Multi-integrated Domain Adaptive Supervision for Fake News Detection | COVID-19 related misinformation and fake news, coined an 'infodemic', has
dramatically increased over the past few years. This misinformation exhibits
concept drift, where the distribution of fake news changes over time, reducing
effectiveness of previously trained models for fake news detection. Given a set
of fake news models trained on multiple domains, we propose an adaptive
decision module to select the best-fit model for a new sample. We propose
MiDAS, a multi-domain adaptative approach for fake news detection that ranks
relevancy of existing models to new samples. MiDAS contains 2 components: a
doman-invariant encoder, and an adaptive model selector. MiDAS integrates
multiple pre-trained and fine-tuned models with their training data to create a
domain-invariant representation. Then, MiDAS uses local Lipschitz smoothness of
the invariant embedding space to estimate each model's relevance to a new
sample. Higher ranked models provide predictions, and lower ranked models
abstain. We evaluate MiDAS on generalization to drifted data with 9 fake news
datasets, each obtained from different domains and modalities. MiDAS achieves
new state-of-the-art performance on multi-domain adaptation for
out-of-distribution fake news classification. |
4,529 | A Novel Weighted Ensemble Learning Based Agent for the Werewolf Game | Werewolf is a popular party game throughout the world, and research on its
significance has progressed in recent years. The Werewolf game is based on
conversation, and in order to win, participants must use all of their cognitive
abilities. This communication game requires the playing agents to be very
sophisticated to win. In this research, we generated a sophisticated agent to
play the Werewolf game using a complex weighted ensemble learning approach.
This research work aimed to estimate what other agents/players think of us in
the game. The agent was developed by aggregating strategies of different
participants in the AI Wolf competition and thereby learning from them using
machine learning. Moreover, the agent created was able to perform much better
than other competitors using very basic strategies to show the approach's
effectiveness in the Werewolf game. The machine learning technique used here is
not restricted to the Werewolf game but may be extended to any game that
requires communication and action depending on other participants. |
4,530 | Calibration Matters: Tackling Maximization Bias in Large-scale Advertising Recommendation Systems | Calibration is defined as the ratio of the average predicted click rate to
the true click rate. The optimization of calibration is essential to many
online advertising recommendation systems because it directly affects the
downstream bids in ads auctions and the amount of money charged to advertisers.
Despite its importance, calibration optimization often suffers from a problem
called "maximization bias". Maximization bias refers to the phenomenon that the
maximum of predicted values overestimates the true maximum. The problem is
introduced because the calibration is computed on the set selected by the
prediction model itself. It persists even if unbiased predictions can be
achieved on every datapoint and worsens when covariate shifts exist between the
training and test sets. To mitigate this problem, we theorize the
quantification of maximization bias and propose a variance-adjusting debiasing
(VAD) meta-algorithm in this paper. The algorithm is efficient, robust, and
practical as it is able to mitigate maximization bias problems under covariate
shifts, neither incurring additional online serving costs nor compromising the
ranking performance. We demonstrate the effectiveness of the proposed algorithm
using a state-of-the-art recommendation neural network model on a large-scale
real-world dataset. |
4,531 | Estimation of Entropy in Constant Space with Improved Sample Complexity | Recent work of Acharya et al. (NeurIPS 2019) showed how to estimate the
entropy of a distribution $\mathcal D$ over an alphabet of size $k$ up to
$\pm\epsilon$ additive error by streaming over $(k/\epsilon^3) \cdot
\text{polylog}(1/\epsilon)$ i.i.d. samples and using only $O(1)$ words of
memory. In this work, we give a new constant memory scheme that reduces the
sample complexity to $(k/\epsilon^2)\cdot \text{polylog}(1/\epsilon)$. We
conjecture that this is optimal up to $\text{polylog}(1/\epsilon)$ factors. |
4,532 | Towards a Holistic View on Argument Quality Prediction | Argumentation is one of society's foundational pillars, and, sparked by
advances in NLP and the vast availability of text data, automated mining of
arguments receives increasing attention. A decisive property of arguments is
their strength or quality. While there are works on the automated estimation of
argument strength, their scope is narrow: they focus on isolated datasets and
neglect the interactions with related argument mining tasks, such as argument
identification, evidence detection, or emotional appeal. In this work, we close
this gap by approaching argument quality estimation from multiple different
angles: Grounded on rich results from thorough empirical evaluations, we assess
the generalization capabilities of argument quality estimation across diverse
domains, the interplay with related argument mining tasks, and the impact of
emotions on perceived argument strength. We find that generalization depends on
a sufficient representation of different domains in the training part. In
zero-shot transfer and multi-task experiments, we reveal that argument quality
is among the more challenging tasks but can improve others. Finally, we show
that emotions play a minor role in argument quality than is often assumed. |
4,533 | Label-invariant Augmentation for Semi-Supervised Graph Classification | Recently, contrastiveness-based augmentation surges a new climax in the
computer vision domain, where some operations, including rotation, crop, and
flip, combined with dedicated algorithms, dramatically increase the model
generalization and robustness. Following this trend, some pioneering attempts
employ the similar idea to graph data. Nevertheless, unlike images, it is much
more difficult to design reasonable augmentations without changing the nature
of graphs. Although exciting, the current graph contrastive learning does not
achieve as promising performance as visual contrastive learning. We conjecture
the current performance of graph contrastive learning might be limited by the
violation of the label-invariant augmentation assumption. In light of this, we
propose a label-invariant augmentation for graph-structured data to address
this challenge. Different from the node/edge modification and subgraph
extraction, we conduct the augmentation in the representation space and
generate the augmented samples in the most difficult direction while keeping
the label of augmented data the same as the original samples. In the
semi-supervised scenario, we demonstrate our proposed method outperforms the
classical graph neural network based methods and recent graph contrastive
learning on eight benchmark graph-structured data, followed by several in-depth
experiments to further explore the label-invariant augmentation in several
aspects. |
4,534 | Graph Neural Networks Are More Powerful Than we Think | Graph Neural Networks (GNNs) are powerful convolutional architectures that
have shown remarkable performance in various node-level and graph-level tasks.
Despite their success, the common belief is that the expressive power of GNNs
is limited and that they are at most as discriminative as the Weisfeiler-Lehman
(WL) algorithm. In this paper we argue the opposite and show that the WL
algorithm is the upper bound only when the input to the GNN is the vector of
all ones. In this direction, we derive an alternative analysis that employs
linear algebraic tools and characterize the representational power of GNNs with
respect to the eigenvalue decomposition of the graph operators. We show that
GNNs can distinguish between any graphs that differ in at least one eigenvalue
and design simple GNN architectures that are provably more expressive than the
WL algorithm. Thorough experimental analysis on graph isomorphism and graph
classification datasets corroborates our theoretical results and demonstrates
the effectiveness of the proposed architectures. |
4,535 | Improving Multi-Task Generalization via Regularizing Spurious Correlation | Multi-Task Learning (MTL) is a powerful learning paradigm to improve
generalization performance via knowledge sharing. However, existing studies
find that MTL could sometimes hurt generalization, especially when two tasks
are less correlated. One possible reason that hurts generalization is spurious
correlation, i.e., some knowledge is spurious and not causally related to task
labels, but the model could mistakenly utilize them and thus fail when such
correlation changes. In MTL setup, there exist several unique challenges of
spurious correlation. First, the risk of having non-causal knowledge is higher,
as the shared MTL model needs to encode all knowledge from different tasks, and
causal knowledge for one task could be potentially spurious to the other.
Second, the confounder between task labels brings in a different type of
spurious correlation to MTL. We theoretically prove that MTL is more prone to
taking non-causal knowledge from other tasks than single-task learning, and
thus generalize worse. To solve this problem, we propose Multi-Task Causal
Representation Learning framework, aiming to represent multi-task knowledge via
disentangled neural modules, and learn which module is causally related to each
task via MTL-specific invariant regularization. Experiments show that it could
enhance MTL model's performance by 5.5% on average over Multi-MNIST, MovieLens,
Taskonomy, CityScape, and NYUv2, via alleviating spurious correlation problem. |
4,536 | Causal Discovery and Injection for Feed-Forward Neural Networks | Neural networks have proven to be effective at solving a wide range of
problems but it is often unclear whether they learn any meaningful causal
relationship: this poses a problem for the robustness of neural network models
and their use for high-stakes decisions. We propose a novel method overcoming
this issue by injecting knowledge in the form of (possibly partial) causal
graphs into feed-forward neural networks, so that the learnt model is
guaranteed to conform to the graph, hence adhering to expert knowledge. This
knowledge may be given up-front or during the learning process, to improve the
model through human-AI collaboration. We apply our method to synthetic and real
(tabular) data showing that it is robust against noise and can improve causal
discovery and prediction performance in low data regimes. |
4,537 | Residual Dynamic Mode Decomposition: Robust and verified Koopmanism | Dynamic Mode Decomposition (DMD) describes complex dynamic processes through
a hierarchy of simpler coherent features. DMD is regularly used to understand
the fundamental characteristics of turbulence and is closely related to Koopman
operators. However, verifying the decomposition, equivalently the computed
spectral features of Koopman operators, remains a major challenge due to the
infinite-dimensional nature of Koopman operators. Challenges include spurious
(unphysical) modes, and dealing with continuous spectra, both of which occur
regularly in turbulent flows. Residual Dynamic Mode Decomposition (ResDMD),
introduced by (Colbrook & Townsend 2021), overcomes some of these challenges
through the data-driven computation of residuals associated with the full
infinite-dimensional Koopman operator. ResDMD computes spectra and
pseudospectra of general Koopman operators with error control, and computes
smoothed approximations of spectral measures (including continuous spectra)
with explicit high-order convergence theorems. ResDMD thus provides robust and
verified Koopmanism. We implement ResDMD and demonstrate its application in a
variety of fluid dynamic situations, at varying Reynolds numbers, arising from
both numerical and experimental data. Examples include: vortex shedding behind
a cylinder; hot-wire data acquired in a turbulent boundary layer; particle
image velocimetry data focusing on a wall-jet flow; and acoustic pressure
signals of laser-induced plasma. We present some advantages of ResDMD, namely,
the ability to verifiably resolve non-linear, transient modes, and spectral
calculation with reduced broadening effects. We also discuss how a new modal
ordering based on residuals enables greater accuracy with a smaller dictionary
than the traditional modulus ordering. This paves the way for greater dynamic
compression of large datasets without sacrificing accuracy. |
4,538 | Understanding Gradient Descent on Edge of Stability in Deep Learning | Deep learning experiments by Cohen et al. [2021] using deterministic Gradient
Descent (GD) revealed an Edge of Stability (EoS) phase when learning rate (LR)
and sharpness (i.e., the largest eigenvalue of Hessian) no longer behave as in
traditional optimization. Sharpness stabilizes around $2/$LR and loss goes up
and down across iterations, yet still with an overall downward trend. The
current paper mathematically analyzes a new mechanism of implicit
regularization in the EoS phase, whereby GD updates due to non-smooth loss
landscape turn out to evolve along some deterministic flow on the manifold of
minimum loss. This is in contrast to many previous results about implicit bias
either relying on infinitesimal updates or noise in gradient. Formally, for any
smooth function $L$ with certain regularity condition, this effect is
demonstrated for (1) Normalized GD, i.e., GD with a varying LR $\eta_t
=\frac{\eta}{|| \nabla L(x(t)) ||}$ and loss $L$; (2) GD with constant LR and
loss $\sqrt{L- \min_x L(x)}$. Both provably enter the Edge of Stability, with
the associated flow on the manifold minimizing $\lambda_{1}(\nabla^2 L)$. The
above theoretical results have been corroborated by an experimental study. |
4,539 | Overcoming Language Disparity in Online Content Classification with Multimodal Learning | Advances in Natural Language Processing (NLP) have revolutionized the way
researchers and practitioners address crucial societal problems. Large language
models are now the standard to develop state-of-the-art solutions for text
detection and classification tasks. However, the development of advanced
computational techniques and resources is disproportionately focused on the
English language, sidelining a majority of the languages spoken globally. While
existing research has developed better multilingual and monolingual language
models to bridge this language disparity between English and non-English
languages, we explore the promise of incorporating the information contained in
images via multimodal machine learning. Our comparative analyses on three
detection tasks focusing on crisis information, fake news, and emotion
recognition, as well as five high-resource non-English languages, demonstrate
that: (a) detection frameworks based on pre-trained large language models like
BERT and multilingual-BERT systematically perform better on the English
language compared against non-English languages, and (b) including images via
multimodal learning bridges this performance gap. We situate our findings with
respect to existing work on the pitfalls of large language models, and discuss
their theoretical and practical implications. Resources for this paper are
available at https://multimodality-language-disparity.github.io/. |
4,540 | Diverse Weight Averaging for Out-of-Distribution Generalization | Standard neural networks struggle to generalize under distribution shifts.
For out-of-distribution generalization in computer vision, the best current
approach averages the weights along a training run. In this paper, we propose
Diverse Weight Averaging (DiWA) that makes a simple change to this strategy:
DiWA averages the weights obtained from several independent training runs
rather than from a single run. Perhaps surprisingly, averaging these weights
performs well under soft constraints despite the network's nonlinearities. The
main motivation behind DiWA is to increase the functional diversity across
averaged models. Indeed, models obtained from different runs are more diverse
than those collected along a single run thanks to differences in
hyperparameters and training procedures. We motivate the need for diversity by
a new bias-variance-covariance-locality decomposition of the expected error,
exploiting similarities between DiWA and standard functional ensembling.
Moreover, this decomposition highlights that DiWA succeeds when the variance
term dominates, which we show happens when the marginal distribution changes at
test time. Experimentally, DiWA consistently improves the state of the art on
the competitive DomainBed benchmark without inference overhead. |
4,541 | Foundation Posteriors for Approximate Probabilistic Inference | Probabilistic programs provide an expressive representation language for
generative models. Given a probabilistic program, we are interested in the task
of posterior inference: estimating a latent variable given a set of observed
variables. Existing techniques for inference in probabilistic programs often
require choosing many hyper-parameters, are computationally expensive, and/or
only work for restricted classes of programs. Here we formulate inference as
masked language modeling: given a program, we generate a supervised dataset of
variables and assignments, and randomly mask a subset of the assignments. We
then train a neural network to unmask the random values, defining an
approximate posterior distribution. By optimizing a single neural network
across a range of programs we amortize the cost of training, yielding a
``foundation'' posterior able to do zero-shot inference for new programs. The
foundation posterior can also be fine-tuned for a particular program and
dataset by optimizing a variational inference objective. We show the efficacy
of the approach, zero-shot and fine-tuned, on a benchmark of STAN programs. |
4,542 | RankGen: Improving Text Generation with Large Ranking Models | Given an input sequence (or prefix), modern language models often assign high
probabilities to output sequences that are repetitive, incoherent, or
irrelevant to the prefix; as such, model-generated text also contains such
artifacts. To address these issues, we present RankGen, an encoder model (1.2B
parameters) that scores model generations given a prefix. RankGen can be
flexibly incorporated as a scoring function in beam search and used to decode
from any pretrained language model. We train RankGen using large-scale
contrastive learning to map a prefix close to the ground-truth sequence that
follows it and far away from two types of negatives: (1) random sequences from
the same document as the prefix, and, which discourage topically-similar but
irrelevant generations; (2) sequences generated from a large language model
conditioned on the prefix, which discourage repetition and hallucination.
Experiments across four different language models (345M-11B parameters) and two
domains show that RankGen significantly outperforms decoding algorithms like
nucleus, top-k, and typical sampling on both automatic metrics (85.0 vs 77.3
MAUVE) as well as human evaluations with English writers (74.5% human
preference over nucleus sampling). Analysis reveals that RankGen outputs are
more relevant to the prefix and improve continuity and coherence compared to
baselines. We open source our model checkpoints, code, and human preferences
with detailed explanations for future research. |
4,543 | Robust and Efficient Medical Imaging with Self-Supervision | Recent progress in Medical Artificial Intelligence (AI) has delivered systems
that can reach clinical expert level performance. However, such systems tend to
demonstrate sub-optimal "out-of-distribution" performance when evaluated in
clinical settings different from the training environment. A common mitigation
strategy is to develop separate systems for each clinical setting using
site-specific data [1]. However, this quickly becomes impractical as medical
data is time-consuming to acquire and expensive to annotate [2]. Thus, the
problem of "data-efficient generalization" presents an ongoing difficulty for
Medical AI development. Although progress in representation learning shows
promise, their benefits have not been rigorously studied, specifically for
out-of-distribution settings. To meet these challenges, we present REMEDIS, a
unified representation learning strategy to improve robustness and
data-efficiency of medical imaging AI. REMEDIS uses a generic combination of
large-scale supervised transfer learning with self-supervised learning and
requires little task-specific customization. We study a diverse range of
medical imaging tasks and simulate three realistic application scenarios using
retrospective data. REMEDIS exhibits significantly improved in-distribution
performance with up to 11.5% relative improvement in diagnostic accuracy over a
strong supervised baseline. More importantly, our strategy leads to strong
data-efficient generalization of medical imaging AI, matching strong supervised
baselines using between 1% to 33% of retraining data across tasks. These
results suggest that REMEDIS can significantly accelerate the life-cycle of
medical imaging AI development thereby presenting an important step forward for
medical imaging AI to deliver broad impact. |
4,544 | HyperAid: Denoising in hyperbolic spaces for tree-fitting and hierarchical clustering | The problem of fitting distances by tree-metrics has received significant
attention in the theoretical computer science and machine learning communities
alike, due to many applications in natural language processing, phylogeny,
cancer genomics and a myriad of problem areas that involve hierarchical
clustering. Despite the existence of several provably exact algorithms for
tree-metric fitting of data that inherently obeys tree-metric constraints, much
less is known about how to best fit tree-metrics for data whose structure
moderately (or substantially) differs from a tree. For such noisy data, most
available algorithms perform poorly and often produce negative edge weights in
representative trees. Furthermore, it is currently not known how to choose the
most suitable approximation objective for noisy fitting. Our contributions are
as follows. First, we propose a new approach to tree-metric denoising
(HyperAid) in hyperbolic spaces which transforms the original data into data
that is ``more'' tree-like, when evaluated in terms of Gromov's $\delta$
hyperbolicity. Second, we perform an ablation study involving two choices for
the approximation objective, $\ell_p$ norms and the Dasgupta loss. Third, we
integrate HyperAid with schemes for enforcing nonnegative edge-weights. As a
result, the HyperAid platform outperforms all other existing methods in the
literature, including Neighbor Joining (NJ), TreeRep and T-REX, both on
synthetic and real-world data. Synthetic data is represented by edge-augmented
trees and shortest-distance metrics while the real-world datasets include Zoo,
Iris, Glass, Segmentation and SpamBase; on these datasets, the average
improvement with respect to NJ is $125.94\%$. |
4,545 | Flexible Modeling and Multitask Learning using Differentiable Tree Ensembles | Decision tree ensembles are widely used and competitive learning models.
Despite their success, popular toolkits for learning tree ensembles have
limited modeling capabilities. For instance, these toolkits support a limited
number of loss functions and are restricted to single task learning. We propose
a flexible framework for learning tree ensembles, which goes beyond existing
toolkits to support arbitrary loss functions, missing responses, and multi-task
learning. Our framework builds on differentiable (a.k.a. soft) tree ensembles,
which can be trained using first-order methods. However, unlike classical
trees, differentiable trees are difficult to scale. We therefore propose a
novel tensor-based formulation of differentiable trees that allows for
efficient vectorization on GPUs. We perform experiments on a collection of 28
real open-source and proprietary datasets, which demonstrate that our framework
can lead to 100x more compact and 23% more expressive tree ensembles than those
by popular toolkits. |
4,546 | Bi-LSTM Scoring Based Similarity Measurement with Agglomerative Hierarchical Clustering (AHC) for Speaker Diarization | Majority of speech signals across different scenarios are never available
with well-defined audio segments containing only a single speaker. A typical
conversation between two speakers consists of segments where their voices
overlap, interrupt each other or halt their speech in between multiple
sentences. Recent advancements in diarization technology leverage neural
network-based approaches to improvise multiple subsystems of speaker
diarization system comprising of extracting segment-wise embedding features and
detecting changes in the speaker during conversation. However, to identify
speaker through clustering, models depend on methodologies like PLDA to
generate similarity measure between two extracted segments from a given
conversational audio. Since these algorithms ignore the temporal structure of
conversations, they tend to achieve a higher Diarization Error Rate (DER), thus
leading to misdetections both in terms of speaker and change identification.
Therefore, to compare similarity of two speech segments both independently and
sequentially, we propose a Bi-directional Long Short-term Memory network for
estimating the elements present in the similarity matrix. Once the similarity
matrix is generated, Agglomerative Hierarchical Clustering (AHC) is applied to
further identify speaker segments based on thresholding. To evaluate the
performance, Diarization Error Rate (DER%) metric is used. The proposed model
achieves a low DER of 34.80% on a test set of audio samples derived from ICSI
Meeting Corpus as compared to traditional PLDA based similarity measurement
mechanism which achieved a DER of 39.90%. |
4,547 | k-strip: A novel segmentation algorithm in k-space for the application of skull stripping | Objectives: Present a novel deep learning-based skull stripping algorithm for
magnetic resonance imaging (MRI) that works directly in the information rich
k-space.
Materials and Methods: Using two datasets from different institutions with a
total of 36,900 MRI slices, we trained a deep learning-based model to work
directly with the complex raw k-space data. Skull stripping performed by HD-BET
(Brain Extraction Tool) in the image domain were used as the ground truth.
Results: Both datasets were very similar to the ground truth (DICE scores of
92\%-98\% and Hausdorff distances of under 5.5 mm). Results on slices above the
eye-region reach DICE scores of up to 99\%, while the accuracy drops in regions
around the eyes and below, with partially blurred output. The output of k-strip
often smoothed edges at the demarcation to the skull. Binary masks are created
with an appropriate threshold.
Conclusion: With this proof-of-concept study, we were able to show the
feasibility of working in the k-space frequency domain, preserving phase
information, with consistent results. Future research should be dedicated to
discovering additional ways the k-space can be used for innovative image
analysis and further workflows. |
4,548 | Extract Dynamic Information To Improve Time Series Modeling: a Case Study with Scientific Workflow | In modeling time series data, we often need to augment the existing data
records to increase the modeling accuracy. In this work, we describe a number
of techniques to extract dynamic information about the current state of a large
scientific workflow, which could be generalized to other types of applications.
The specific task to be modeled is the time needed for transferring a file from
an experimental facility to a data center. The key idea of our approach is to
find recent past data transfer events that match the current event in some
ways. Tests showed that we could identify recent events matching some recorded
properties and reduce the prediction error by about 12% compared to the similar
models with only static features. We additionally explored an application
specific technique to extract information about the data production process,
and was able to reduce the average prediction error by 44%. |
4,549 | Parallel and Distributed Graph Neural Networks: An In-Depth Concurrency Analysis | Graph neural networks (GNNs) are among the most powerful tools in deep
learning. They routinely solve complex problems on unstructured networks, such
as node classification, graph classification, or link prediction, with high
accuracy. However, both inference and training of GNNs are complex, and they
uniquely combine the features of irregular graph processing with dense and
regular computations. This complexity makes it very challenging to execute GNNs
efficiently on modern massively parallel architectures. To alleviate this, we
first design a taxonomy of parallelism in GNNs, considering data and model
parallelism, and different forms of pipelining. Then, we use this taxonomy to
investigate the amount of parallelism in numerous GNN models, GNN-driven
machine learning tasks, software frameworks, or hardware accelerators. We use
the work-depth model, and we also assess communication volume and
synchronization. We specifically focus on the sparsity/density of the
associated tensors, in order to understand how to effectively apply techniques
such as vectorization. We also formally analyze GNN pipelining, and we
generalize the established Message-Passing class of GNN models to cover
arbitrary pipeline depths, facilitating future optimizations. Finally, we
investigate different forms of asynchronicity, navigating the path for future
asynchronous parallel GNN pipelines. The outcomes of our analysis are
synthesized in a set of insights that help to maximize GNN performance, and a
comprehensive list of challenges and opportunities for further research into
efficient GNN computations. Our work will help to advance the design of future
GNNs. |
4,550 | Neural network topological snake models for locating general phase diagrams | Machine learning for locating phase diagram has received intensive research
interest in recent years. However, its application in automatically locating
phase diagram is limited to single closed phase boundary. In this paper, in
order to locate phase diagrams with multiple phases and complex boundaries, we
introduce (i) a network-shaped snake model and (ii) a topologically
transformable snake with discriminative cooperative networks, respectively. The
phase diagrams of both quantum and classical spin-1 model are obtained. Our
method is flexible to determine the phase diagram with just snapshots of
configurations from the cold-atom or other experiments. |
4,551 | ArabGlossBERT: Fine-Tuning BERT on Context-Gloss Pairs for WSD | Using pre-trained transformer models such as BERT has proven to be effective
in many NLP tasks. This paper presents our work to fine-tune BERT models for
Arabic Word Sense Disambiguation (WSD). We treated the WSD task as a
sentence-pair binary classification task. First, we constructed a dataset of
labeled Arabic context-gloss pairs (~167k pairs) we extracted from the Arabic
Ontology and the large lexicographic database available at Birzeit University.
Each pair was labeled as True or False and target words in each context were
identified and annotated. Second, we used this dataset for fine-tuning three
pre-trained Arabic BERT models. Third, we experimented the use of different
supervised signals used to emphasize target words in context. Our experiments
achieved promising results (accuracy of 84%) although we used a large set of
senses in the experiment. |
4,552 | Dexterous Robotic Manipulation using Deep Reinforcement Learning and Knowledge Transfer for Complex Sparse Reward-based Tasks | This paper describes a deep reinforcement learning (DRL) approach that won
Phase 1 of the Real Robot Challenge (RRC) 2021, and then extends this method to
a more difficult manipulation task. The RRC consisted of using a TriFinger
robot to manipulate a cube along a specified positional trajectory, but with no
requirement for the cube to have any specific orientation. We used a relatively
simple reward function, a combination of goal-based sparse reward and distance
reward, in conjunction with Hindsight Experience Replay (HER) to guide the
learning of the DRL agent (Deep Deterministic Policy Gradient (DDPG)). Our
approach allowed our agents to acquire dexterous robotic manipulation
strategies in simulation. These strategies were then applied to the real robot
and outperformed all other competition submissions, including those using more
traditional robotic control techniques, in the final evaluation stage of the
RRC. Here we extend this method, by modifying the task of Phase 1 of the RRC to
require the robot to maintain the cube in a particular orientation, while the
cube is moved along the required positional trajectory. The requirement to also
orient the cube makes the agent unable to learn the task through blind
exploration due to increased problem complexity. To circumvent this issue, we
make novel use of a Knowledge Transfer (KT) technique that allows the
strategies learned by the agent in the original task (which was agnostic to
cube orientation) to be transferred to this task (where orientation matters).
KT allowed the agent to learn and perform the extended task in the simulator,
which improved the average positional deviation from 0.134 m to 0.02 m, and
average orientation deviation from 142{\deg} to 76{\deg} during evaluation.
This KT concept shows good generalisation properties and could be applied to
any actor-critic learning algorithm. |
4,553 | Metrics of calibration for probabilistic predictions | Predictions are often probabilities; e.g., a prediction could be for
precipitation tomorrow, but with only a 30% chance. Given such probabilistic
predictions together with the actual outcomes, "reliability diagrams" help
detect and diagnose statistically significant discrepancies -- so-called
"miscalibration" -- between the predictions and the outcomes. The canonical
reliability diagrams histogram the observed and expected values of the
predictions; replacing the hard histogram binning with soft kernel density
estimation is another common practice. But, which widths of bins or kernels are
best? Plots of the cumulative differences between the observed and expected
values largely avoid this question, by displaying miscalibration directly as
the slopes of secant lines for the graphs. Slope is easy to perceive with
quantitative precision, even when the constant offsets of the secant lines are
irrelevant; there is no need to bin or perform kernel density estimation.
The existing standard metrics of miscalibration each summarize a reliability
diagram as a single scalar statistic. The cumulative plots naturally lead to
scalar metrics for the deviation of the graph of cumulative differences away
from zero; good calibration corresponds to a horizontal, flat graph which
deviates little from zero. The cumulative approach is currently unconventional,
yet offers many favorable statistical properties, guaranteed via mathematical
theory backed by rigorous proofs and illustrative numerical examples. In
particular, metrics based on binning or kernel density estimation unavoidably
must trade-off statistical confidence for the ability to resolve variations as
a function of the predicted probability or vice versa. Widening the bins or
kernels averages away random noise while giving up some resolving power.
Narrowing the bins or kernels enhances resolving power while not averaging away
as much noise. |
4,554 | Semi-Supervised Learning for Image Classification using Compact Networks in the BioMedical Context | The development of mobile and on the edge applications that embed deep
convolutional neural models has the potential to revolutionise biomedicine.
However, most deep learning models require computational resources that are not
available in smartphones or edge devices; an issue that can be faced by means
of compact models. The problem with such models is that they are, at least
usually, less accurate than bigger models. In this work, we study how this
limitation can be addressed with the application of semi-supervised learning
techniques. We conduct several statistical analyses to compare performance of
deep compact architectures when trained using semi-supervised learning methods
for tackling image classification tasks in the biomedical context. In
particular, we explore three families of compact networks, and two families of
semi-supervised learning techniques for 10 biomedical tasks. By combining
semi-supervised learning methods with compact networks, it is possible to
obtain a similar performance to standard size networks. In general, the best
results are obtained when combining data distillation with MixNet, and plain
distillation with ResNet-18. Also, in general, NAS networks obtain better
results than manually designed networks and quantized networks. The work
presented in this paper shows the benefits of apply semi-supervised methods to
compact networks; this allow us to create compact models that are not only as
accurate as standard size models, but also faster and lighter. Finally, we have
developed a library that simplifies the construction of compact models using
semi-supervised learning methods. |
4,555 | Beyond Greedy Search: Tracking by Multi-Agent Reinforcement Learning-based Beam Search | To track the target in a video, current visual trackers usually adopt greedy
search for target object localization in each frame, that is, the candidate
region with the maximum response score will be selected as the tracking result
of each frame. However, we found that this may be not an optimal choice,
especially when encountering challenging tracking scenarios such as heavy
occlusion and fast motion. To address this issue, we propose to maintain
multiple tracking trajectories and apply beam search strategy for visual
tracking, so that the trajectory with fewer accumulated errors can be
identified. Accordingly, this paper introduces a novel multi-agent
reinforcement learning based beam search tracking strategy, termed
BeamTracking. It is mainly inspired by the image captioning task, which takes
an image as input and generates diverse descriptions using beam search
algorithm. Accordingly, we formulate the tracking as a sample selection problem
fulfilled by multiple parallel decision-making processes, each of which aims at
picking out one sample as their tracking result in each frame. Each maintained
trajectory is associated with an agent to perform the decision-making and
determine what actions should be taken to update related information. When all
the frames are processed, we select the trajectory with the maximum accumulated
score as the tracking result. Extensive experiments on seven popular tracking
benchmark datasets validated the effectiveness of the proposed algorithm. |
4,556 | Disentangling Active and Passive Cosponsorship in the U.S. Congress | In the U.S. Congress, legislators can use active and passive cosponsorship to
support bills. We show that these two types of cosponsorship are driven by two
different motivations: the backing of political colleagues and the backing of
the bill's content. To this end, we develop an Encoder+RGCN based model that
learns legislator representations from bill texts and speech transcripts. These
representations predict active and passive cosponsorship with an F1-score of
0.88. Applying our representations to predict voting decisions, we show that
they are interpretable and generalize to unseen tasks. |
4,557 | Detect Professional Malicious User with Metric Learning in Recommender Systems | In e-commerce, online retailers are usually suffering from professional
malicious users (PMUs), who utilize negative reviews and low ratings to their
consumed products on purpose to threaten the retailers for illegal profits.
Specifically, there are three challenges for PMU detection: 1) professional
malicious users do not conduct any abnormal or illegal interactions (they never
concurrently leave too many negative reviews and low ratings at the same time),
and they conduct masking strategies to disguise themselves. Therefore,
conventional outlier detection methods are confused by their masking
strategies. 2) the PMU detection model should take both ratings and reviews
into consideration, which makes PMU detection a multi-modal problem. 3) there
are no datasets with labels for professional malicious users in public, which
makes PMU detection an unsupervised learning problem. To this end, we propose
an unsupervised multi-modal learning model: MMD, which employs Metric learning
for professional Malicious users Detection with both ratings and reviews. MMD
first utilizes a modified RNN to project the informational review into a
sentiment score, which jointly considers the ratings and reviews. Then
professional malicious user profiling (MUP) is proposed to catch the sentiment
gap between sentiment scores and ratings. MUP filters the users and builds a
candidate PMU set. We apply a metric learning-based clustering to learn a
proper metric matrix for PMU detection. Finally, we can utilize this metric and
labeled users to detect PMUs. Specifically, we apply the attention mechanism in
metric learning to improve the model's performance. The extensive experiments
in four datasets demonstrate that our proposed method can solve this
unsupervised detection problem. Moreover, the performance of the
state-of-the-art recommender models is enhanced by taking MMD as a
preprocessing stage. |
4,558 | Semi-WTC: A Practical Semi-supervised Framework for Attack Categorization through Weight-Task Consistency | Supervised learning has been widely used for attack detection, which requires
large amounts of high-quality data and labels. However, the data is often
imbalanced and sufficient annotations are difficult to obtain. Moreover, these
supervised models are subject to real-world deployment issues, such as
defending against unseen artificial attacks. We propose a semi-supervised
fine-grained attack categorization framework consisting of an encoder and a
two-branch structure to integrate information from labeled and unlabeled data
to tackle these practical challenges. This framework can be generalized to
different supervised models. The multilayer perceptron with residual connection
and batch normalization is used as the encoder to extract features and reduce
the complexity. The Recurrent Prototype Module (RPM) is proposed to train the
encoder effectively in a semi-supervised manner. To alleviate the problem of
data imbalance, we introduce the Weight-Task Consistency (WTC) into the
iterative process of RPM by assigning larger weights to classes with fewer
samples in the loss function. In addition, to cope with new attacks in
real-world deployment, we further propose an Active Adaption Resampling (AAR)
method, which can better discover the distribution of the unseen sample data
and adapt the parameters of the encoder. Experimental results show that our
model outperforms the state-of-the-art semi-supervised attack detection methods
with a general 5% improvement in classification accuracy and a 90% reduction in
training time. |
4,559 | The AI Mechanic: Acoustic Vehicle Characterization Neural Networks | In a world increasingly dependent on road-based transportation, it is
essential to understand vehicles. We introduce the AI mechanic, an acoustic
vehicle characterization deep learning system, as an integrated approach using
sound captured from mobile devices to enhance transparency and understanding of
vehicles and their condition for non-expert users. We develop and implement
novel cascading architectures for vehicle understanding, which we define as
sequential, conditional, multi-level networks that process raw audio to extract
highly-granular insights. To showcase the viability of cascading architectures,
we build a multi-task convolutional neural network that predicts and cascades
vehicle attributes to enhance fault detection. We train and test these models
on a synthesized dataset reflecting more than 40 hours of augmented audio and
achieve >92% validation set accuracy on attributes (fuel type, engine
configuration, cylinder count and aspiration type). Our cascading architecture
additionally achieved 93.6% validation and 86.8% test set accuracy on misfire
fault prediction, demonstrating margins of 16.4% / 7.8% and 4.2% / 1.5%
improvement over na\"ive and parallel baselines. We explore experimental
studies focused on acoustic features, data augmentation, feature fusion, and
data reliability. Finally, we conclude with a discussion of broader
implications, future directions, and application areas for this work. |
4,560 | Self-Consistent Dynamical Field Theory of Kernel Evolution in Wide Neural Networks | We analyze feature learning in infinite width neural networks trained with
gradient flow through a self-consistent dynamical field theory. We construct a
collection of deterministic dynamical order parameters which are inner-product
kernels for hidden unit activations and gradients in each layer at pairs of
time points, providing a reduced description of network activity through
training. These kernel order parameters collectively define the hidden layer
activation distribution, the evolution of the neural tangent kernel, and
consequently output predictions. For deep linear networks, these kernels
satisfy a set of algebraic matrix equations. For nonlinear networks, we provide
an alternating sampling procedure to self-consistently solve for the kernel
order parameters. We provide comparisons of the self-consistent solution to
various approximation schemes including the static NTK approximation, gradient
independence assumption, and leading order perturbation theory, showing that
each of these approximations can break down in regimes where general
self-consistent solutions still provide an accurate description. Lastly, we
provide experiments in more realistic settings which demonstrate that the loss
and kernel dynamics of CNNs at fixed feature learning strength is preserved
across different widths on a CIFAR classification task. |
4,561 | Wojood: Nested Arabic Named Entity Corpus and Recognition using BERT | This paper presents Wojood, a corpus for Arabic nested Named Entity
Recognition (NER). Nested entities occur when one entity mention is embedded
inside another entity mention. Wojood consists of about 550K Modern Standard
Arabic (MSA) and dialect tokens that are manually annotated with 21 entity
types including person, organization, location, event and date. More
importantly, the corpus is annotated with nested entities instead of the more
common flat annotations. The data contains about 75K entities and 22.5% of
which are nested. The inter-annotator evaluation of the corpus demonstrated a
strong agreement with Cohen's Kappa of 0.979 and an F1-score of 0.976. To
validate our data, we used the corpus to train a nested NER model based on
multi-task learning and AraBERT (Arabic BERT). The model achieved an overall
micro F1-score of 0.884. Our corpus, the annotation guidelines, the source code
and the pre-trained model are publicly available. |
4,562 | Are Graph Representation Learning Methods Robust to Graph Sparsity and Asymmetric Node Information? | The growing popularity of Graph Representation Learning (GRL) methods has
resulted in the development of a large number of models applied to a miscellany
of domains. Behind this diversity of domains, there is a strong heterogeneity
of graphs, making it difficult to estimate the expected performance of a model
on a new graph, especially when the graph has distinctive characteristics that
have not been encountered in the benchmark yet. To address this, we have
developed an experimental pipeline, to assess the impact of a given property on
the models performances. In this paper, we use this pipeline to study the
effect of two specificities encountered on banks transactional graphs resulting
from the partial view a bank has on all the individuals and transactions
carried out on the market. These specific features are graph sparsity and
asymmetric node information. This study demonstrates the robustness of GRL
methods to these distinctive characteristics. We believe that this work can
ease the evaluation of GRL methods to specific characteristics and foster the
development of such methods on transactional graphs. |
4,563 | The First Optimal Acceleration of High-Order Methods in Smooth Convex Optimization | In this paper, we study the fundamental open question of finding the optimal
high-order algorithm for solving smooth convex minimization problems. Arjevani
et al. (2019) established the lower bound
$\Omega\left(\epsilon^{-2/(3p+1)}\right)$ on the number of the $p$-th order
oracle calls required by an algorithm to find an $\epsilon$-accurate solution
to the problem, where the $p$-th order oracle stands for the computation of the
objective function value and the derivatives up to the order $p$. However, the
existing state-of-the-art high-order methods of Gasnikov et al. (2019b); Bubeck
et al. (2019); Jiang et al. (2019) achieve the oracle complexity
$\mathcal{O}\left(\epsilon^{-2/(3p+1)} \log (1/\epsilon)\right)$, which does
not match the lower bound. The reason for this is that these algorithms require
performing a complex binary search procedure, which makes them neither optimal
nor practical. We fix this fundamental issue by providing the first algorithm
with $\mathcal{O}\left(\epsilon^{-2/(3p+1)}\right)$ $p$-th order oracle
complexity. |
4,564 | Certified Error Control of Candidate Set Pruning for Two-Stage Relevance Ranking | In information retrieval (IR), candidate set pruning has been commonly used
to speed up two-stage relevance ranking. However, such an approach lacks
accurate error control and often trades accuracy off against computational
efficiency in an empirical fashion, lacking theoretical guarantees. In this
paper, we propose the concept of certified error control of candidate set
pruning for relevance ranking, which means that the test error after pruning is
guaranteed to be controlled under a user-specified threshold with high
probability. Both in-domain and out-of-domain experiments show that our method
successfully prunes the first-stage retrieved candidate sets to improve the
second-stage reranking speed while satisfying the pre-specified accuracy
constraints in both settings. For example, on MS MARCO Passage v1, our method
yields an average candidate set size of 27 out of 1,000 which increases the
reranking speed by about 37 times, while the MRR@10 is greater than a
pre-specified value of 0.38 with about 90% empirical coverage and the empirical
baselines fail to provide such guarantee. Code and data are available at:
https://github.com/alexlimh/CEC-Ranking. |
4,565 | What killed the Convex Booster ? | A landmark negative result of Long and Servedio established a worst-case
spectacular failure of a supervised learning trio (loss, algorithm, model)
otherwise praised for its high precision machinery. Hundreds of papers followed
up on the two suspected culprits: the loss (for being convex) and/or the
algorithm (for fitting a classical boosting blueprint). Here, we call to the
half-century+ founding theory of losses for class probability estimation
(properness), an extension of Long and Servedio's results and a new general
boosting algorithm to demonstrate that the real culprit in their specific
context was in fact the (linear) model class. We advocate for a more general
stanpoint on the problem as we argue that the source of the negative result
lies in the dark side of a pervasive -- and otherwise prized -- aspect of ML:
\textit{parameterisation}. |
4,566 | Focused Adversarial Attacks | Recent advances in machine learning show that neural models are vulnerable to
minimally perturbed inputs, or adversarial examples. Adversarial algorithms are
optimization problems that minimize the accuracy of ML models by perturbing
inputs, often using a model's loss function to craft such perturbations.
State-of-the-art object detection models are characterized by very large output
manifolds due to the number of possible locations and sizes of objects in an
image. This leads to their outputs being sparse and optimization problems that
use them incur a lot of unnecessary computation.
We propose to use a very limited subset of a model's learned manifold to
compute adversarial examples. Our \textit{Focused Adversarial Attacks} (FA)
algorithm identifies a small subset of sensitive regions to perform
gradient-based adversarial attacks. FA is significantly faster than other
gradient-based attacks when a model's manifold is sparsely activated. Also, its
perturbations are more efficient than other methods under the same perturbation
constraints. We evaluate FA on the COCO 2017 and Pascal VOC 2007 detection
datasets. |
4,567 | What Is Fairness? Implications For FairML | A growing body of literature in fairness-aware ML (fairML) aspires to
mitigate machine learning (ML)-related unfairness in automated decision making
(ADM) by defining metrics that measure fairness of an ML model and by proposing
methods that ensure that trained ML models achieve low values in those
measures. However, the underlying concept of fairness, i.e., the question of
what fairness is, is rarely discussed, leaving a considerable gap between
centuries of philosophical discussion and recent adoption of the concept in the
ML community. In this work, we try to bridge this gap by formalizing a
consistent concept of fairness and by translating the philosophical
considerations into a formal framework for the evaluation of ML models in ADM
systems. We derive that fairness problems can already arise without the
presence of protected attributes, pointing out that fairness and predictive
performance are not irreconcilable counterparts, but rather that the latter is
necessary to achieve the former. Moreover, we argue why and how causal
considerations are necessary when assessing fairness in the presence of
protected attributes. Eventually, we achieve greater linguistic clarity for the
discussion of fairML by clearly assigning responsibilities to stakeholders
inside and outside ML. |
4,568 | Towards a Theory of Faithfulness: Faithful Explanations of Differentiable Classifiers over Continuous Data | There is broad agreement in the literature that explanation methods should be
faithful to the model that they explain, but faithfulness remains a rather
vague term. We revisit faithfulness in the context of continuous data and
propose two formal definitions of faithfulness for feature attribution methods.
Qualitative faithfulness demands that scores reflect the true qualitative
effect (positive vs. negative) of the feature on the model and quanitative
faithfulness that the magnitude of scores reflect the true quantitative effect.
We discuss under which conditions these requirements can be satisfied to which
extent (local vs global). As an application of the conceptual idea, we look at
differentiable classifiers over continuous data and characterize
Gradient-scores as follows: every qualitatively faithful feature attribution
method is qualitatively equivalent to Gradient-scores. Furthermore, if an
attribution method is quantitatively faithful in the sense that changes of the
output of the classifier are proportional to the scores of features, then it is
either equivalent to gradient-scoring or it is based on an inferior
approximation of the classifier. To illustrate the practical relevance of the
theory, we experimentally demonstrate that popular attribution methods can fail
to give faithful explanations in the setting where the data is continuous and
the classifier differentiable. |
4,569 | Improving Robustness against Real-World and Worst-Case Distribution Shifts through Decision Region Quantification | The reliability of neural networks is essential for their use in
safety-critical applications. Existing approaches generally aim at improving
the robustness of neural networks to either real-world distribution shifts
(e.g., common corruptions and perturbations, spatial transformations, and
natural adversarial examples) or worst-case distribution shifts (e.g.,
optimized adversarial examples). In this work, we propose the Decision Region
Quantification (DRQ) algorithm to improve the robustness of any differentiable
pre-trained model against both real-world and worst-case distribution shifts in
the data. DRQ analyzes the robustness of local decision regions in the vicinity
of a given data point to make more reliable predictions. We theoretically
motivate the DRQ algorithm by showing that it effectively smooths spurious
local extrema in the decision surface. Furthermore, we propose an
implementation using targeted and untargeted adversarial attacks. An extensive
empirical evaluation shows that DRQ increases the robustness of adversarially
and non-adversarially trained models against real-world and worst-case
distribution shifts on several computer vision benchmark datasets. |
4,570 | A Topological Approach for Semi-Supervised Learning | Nowadays, Machine Learning and Deep Learning methods have become the
state-of-the-art approach to solve data classification tasks. In order to use
those methods, it is necessary to acquire and label a considerable amount of
data; however, this is not straightforward in some fields, since data
annotation is time consuming and might require expert knowledge. This challenge
can be tackled by means of semi-supervised learning methods that take advantage
of both labelled and unlabelled data. In this work, we present new
semi-supervised learning methods based on techniques from Topological Data
Analysis (TDA), a field that is gaining importance for analysing large amounts
of data with high variety and dimensionality. In particular, we have created
two semi-supervised learning methods following two different topological
approaches. In the former, we have used a homological approach that consists in
studying the persistence diagrams associated with the data using the Bottleneck
and Wasserstein distances. In the latter, we have taken into account the
connectivity of the data. In addition, we have carried out a thorough analysis
of the developed methods using 3 synthetic datasets, 5 structured datasets, and
2 datasets of images. The results show that the semi-supervised methods
developed in this work outperform both the results obtained with models trained
with only manually labelled data, and those obtained with classical
semi-supervised learning methods, reaching improvements of up to a 16%. |
4,571 | Women, artificial intelligence, and key positions in collaboration networks: Towards a more equal scientific ecosystem | Scientific collaboration in almost every discipline is mainly driven by the
need of sharing knowledge, expertise, and pooled resources. Science is becoming
more complex which has encouraged scientists to involve more in collaborative
research projects in order to better address the challenges. As a highly
interdisciplinary field with a rapidly evolving scientific landscape,
artificial intelligence calls for researchers with special profiles covering a
diverse set of skills and expertise. Understanding gender aspects of scientific
collaboration is of paramount importance, especially in a field such as
artificial intelligence that has been attracting large investments. Using
social network analysis, natural language processing, and machine learning and
focusing on artificial intelligence publications for the period from 2000 to
2019, in this work, we comprehensively investigated the effects of several
driving factors on acquiring key positions in scientific collaboration networks
through a gender lens. It was found that, regardless of gender, scientific
performance in terms of quantity and impact plays a crucial in possessing the
"social researcher" in the network. However, subtle differences were observed
between female and male researchers in acquiring the "local influencer" role. |
4,572 | EXACT: How to Train Your Accuracy | Classification tasks are usually evaluated in terms of accuracy. However,
accuracy is discontinuous and cannot be directly optimized using gradient
ascent. Popular methods minimize cross-entropy, Hinge loss, or other surrogate
losses, which can lead to suboptimal results. In this paper, we propose a new
optimization framework by introducing stochasticity to a model's output and
optimizing expected accuracy, i.e. accuracy of the stochastic model. Extensive
experiments on image classification show that the proposed optimization method
is a powerful alternative to widely used classification losses. |
4,573 | CLCNet: Rethinking of Ensemble Modeling with Classification Confidence Network | In this paper, we propose a Classification Confidence Network (CLCNet) that
can determine whether the classification model classifies input samples
correctly. It can take a classification result in the form of vector in any
dimension, and return a confidence score as output, which represents the
probability of an instance being classified correctly. We can utilize CLCNet in
a simple cascade structure system consisting of several SOTA (state-of-the-art)
classification models, and our experiments show that the system can achieve the
following advantages: 1. The system can customize the average computation
requirement (FLOPs) per image while inference. 2. Under the same computation
requirement, the performance of the system can exceed any model that has
identical structure with the model in the system, but different in size. In
fact, this is a new type of ensemble modeling. Like general ensemble modeling,
it can achieve higher performance than single classification model, yet our
system requires much less computation than general ensemble modeling. We have
uploaded our code to a github repository:
https://github.com/yaoching0/CLCNet-Rethinking-of-Ensemble-Modeling. |
4,574 | Learning Energy Networks with Generalized Fenchel-Young Losses | Energy-based models, a.k.a. energy networks, perform inference by optimizing
an energy function, typically parametrized by a neural network. This allows one
to capture potentially complex relationships between inputs and outputs. To
learn the parameters of the energy function, the solution to that optimization
problem is typically fed into a loss function. The key challenge for training
energy networks lies in computing loss gradients, as this typically requires
argmin/argmax differentiation. In this paper, building upon a generalized
notion of conjugate function, which replaces the usual bilinear pairing with a
general energy function, we propose generalized Fenchel-Young losses, a natural
loss construction for learning energy networks. Our losses enjoy many desirable
properties and their gradients can be computed efficiently without
argmin/argmax differentiation. We also prove the calibration of their excess
risk in the case of linear-concave energies. We demonstrate our losses on
multilabel classification and imitation learning tasks. |
4,575 | How catastrophic can catastrophic forgetting be in linear regression? | To better understand catastrophic forgetting, we study fitting an
overparameterized linear model to a sequence of tasks with different input
distributions. We analyze how much the model forgets the true labels of earlier
tasks after training on subsequent tasks, obtaining exact expressions and
bounds. We establish connections between continual learning in the linear
setting and two other research areas: alternating projections and the Kaczmarz
method. In specific settings, we highlight differences between forgetting and
convergence to the offline solution as studied in those areas. In particular,
when T tasks in d dimensions are presented cyclically for k iterations, we
prove an upper bound of T^2 * min{1/sqrt(k), d/k} on the forgetting. This
stands in contrast to the convergence to the offline solution, which can be
arbitrarily slow according to existing alternating projection results. We
further show that the T^2 factor can be lifted when tasks are presented in a
random ordering. |
4,576 | Discovering Dynamic Functional Brain Networks via Spatial and Channel-wise Attention | Using deep learning models to recognize functional brain networks (FBNs) in
functional magnetic resonance imaging (fMRI) has been attracting increasing
interest recently. However, most existing work focuses on detecting static FBNs
from entire fMRI signals, such as correlation-based functional connectivity.
Sliding-window is a widely used strategy to capture the dynamics of FBNs, but
it is still limited in representing intrinsic functional interactive dynamics
at each time step. And the number of FBNs usually need to be set manually. More
over, due to the complexity of dynamic interactions in brain, traditional
linear and shallow models are insufficient in identifying complex and spatially
overlapped FBNs across each time step. In this paper, we propose a novel
Spatial and Channel-wise Attention Autoencoder (SCAAE) for discovering FBNs
dynamically. The core idea of SCAAE is to apply attention mechanism to FBNs
construction. Specifically, we designed two attention modules: 1) spatial-wise
attention (SA) module to discover FBNs in the spatial domain and 2) a
channel-wise attention (CA) module to weigh the channels for selecting the FBNs
automatically. We evaluated our approach on ADHD200 dataset and our results
indicate that the proposed SCAAE method can effectively recover the dynamic
changes of the FBNs at each fMRI time step, without using sliding windows. More
importantly, our proposed hybrid attention modules (SA and CA) do not enforce
assumptions of linearity and independence as previous methods, and thus provide
a novel approach to better understanding dynamic functional brain networks. |
4,577 | Learning Graph Structure from Convolutional Mixtures | Machine learning frameworks such as graph neural networks typically rely on a
given, fixed graph to exploit relational inductive biases and thus effectively
learn from network data. However, when said graphs are (partially) unobserved,
noisy, or dynamic, the problem of inferring graph structure from data becomes
relevant. In this paper, we postulate a graph convolutional relationship
between the observed and latent graphs, and formulate the graph learning task
as a network inverse (deconvolution) problem. In lieu of
eigendecomposition-based spectral methods or iterative optimization solutions,
we unroll and truncate proximal gradient iterations to arrive at a
parameterized neural network architecture that we call a Graph Deconvolution
Network (GDN). GDNs can learn a distribution of graphs in a supervised fashion,
perform link prediction or edge-weight regression tasks by adapting the loss
function, and they are inherently inductive. We corroborate GDN's superior
graph recovery performance and its generalization to larger graphs using
synthetic data in supervised settings. Furthermore, we demonstrate the
robustness and representation power of GDNs on real world neuroimaging and
social network datasets. |
4,578 | Jacobian Granger Causal Neural Networks for Analysis of Stationary and Nonstationary Data | Granger causality is a commonly used method for uncovering information flow
and dependencies in a time series. Here we introduce JGC (Jacobian Granger
Causality), a neural network-based approach to Granger causality using the
Jacobian as a measure of variable importance, and propose a thresholding
procedure for inferring Granger causal variables using this measure. The
resulting approach performs consistently well compared to other approaches in
identifying Granger causal variables, the associated time lags, as well as
interaction signs. Lastly, through the inclusion of a time variable, we show
that this approach is able to learn the temporal dependencies for nonstationary
systems whose Granger causal structures change in time. |
4,579 | Provably Precise, Succinct and Efficient Explanations for Decision Trees | Decision trees (DTs) embody interpretable classifiers. DTs have been
advocated for deployment in high-risk applications, but also for explaining
other complex classifiers. Nevertheless, recent work has demonstrated that
predictions in DTs ought to be explained with rigorous approaches. Although
rigorous explanations can be computed in polynomial time for DTs, their size
may be beyond the cognitive limits of human decision makers. This paper
investigates the computation of {\delta}-relevant sets for DTs.
{\delta}-relevant sets denote explanations that are succinct and provably
precise. These sets represent generalizations of rigorous explanations, which
are precise with probability one, and so they enable trading off explanation
size for precision. The paper proposes two logic encodings for computing
smallest {\delta}-relevant sets for DTs. The paper further devises a
polynomial-time algorithm for computing {\delta}-relevant sets which are not
guaranteed to be subset-minimal, but for which the experiments show to be most
often subset-minimal in practice. The experimental results also demonstrate the
practical efficiency of computing smallest {\delta}-relevant sets. |
4,580 | Automatic Spoken Language Identification using a Time-Delay Neural Network | Closed-set spoken language identification is the task of recognizing the
language being spoken in a recorded audio clip from a set of known languages.
In this study, a language identification system was built and trained to
distinguish between Arabic, Spanish, French, and Turkish based on nothing more
than recorded speech. A pre-existing multilingual dataset was used to train a
series of acoustic models based on the Tedlium TDNN model to perform automatic
speech recognition. The system was provided with a custom multilingual language
model and a specialized pronunciation lexicon with language names prepended to
phones. The trained model was used to generate phone alignments to test data
from all four languages, and languages were predicted based on a voting scheme
choosing the most common language prepend in an utterance. Accuracy was
measured by comparing predicted languages to known languages, and was
determined to be very high in identifying Spanish and Arabic, and somewhat
lower in identifying Turkish and French. |
4,581 | Hybrid Intelligent Testing in Simulation-Based Verification | Efficient and effective testing for simulation-based hardware verification is
challenging. Using constrained random test generation, several millions of
tests may be required to achieve coverage goals. The vast majority of tests do
not contribute to coverage progress, yet they consume verification resources.
In this paper, we propose a hybrid intelligent testing approach combining two
methods that have previously been treated separately, namely Coverage-Directed
Test Selection and Novelty-Driven Verification. Coverage-Directed Test
Selection learns from coverage feedback to bias testing towards the most
effective tests. Novelty-Driven Verification learns to identify and simulate
stimuli that differ from previous stimuli, thereby reducing the number of
simulations and increasing testing efficiency. We discuss the strengths and
limitations of each method, and we show how our approach addresses each
method's limitations, leading to hardware testing that is both efficient and
effective. |
4,582 | Data Valuation for Offline Reinforcement Learning | The success of deep reinforcement learning (DRL) hinges on the availability
of training data, which is typically obtained via a large number of environment
interactions. In many real-world scenarios, costs and risks are associated with
gathering these data. The field of offline reinforcement learning addresses
these issues through outsourcing the collection of data to a domain expert or a
carefully monitored program and subsequently searching for a batch-constrained
optimal policy. With the emergence of data markets, an alternative to
constructing a dataset in-house is to purchase external data. However, while
state-of-the-art offline reinforcement learning approaches have shown a lot of
promise, they currently rely on carefully constructed datasets that are well
aligned with the intended target domains. This raises questions regarding the
transferability and robustness of an offline reinforcement learning agent
trained on externally acquired data. In this paper, we empirically evaluate the
ability of the current state-of-the-art offline reinforcement learning
approaches to coping with the source-target domain mismatch within two MuJoCo
environments, finding that current state-of-the-art offline reinforcement
learning algorithms underperform in the target domain. To address this, we
propose data valuation for offline reinforcement learning (DVORL), which allows
us to identify relevant and high-quality transitions, improving the performance
and transferability of policies learned by offline reinforcement learning
algorithms. The results show that our method outperforms offline reinforcement
learning baselines on two MuJoCo environments. |
4,583 | ODBO: Bayesian Optimization with Search Space Prescreening for Directed Protein Evolution | Directed evolution is a versatile technique in protein engineering that
mimics the process of natural selection by iteratively alternating between
mutagenesis and screening in order to search for sequences that optimize a
given property of interest, such as catalytic activity and binding affinity to
a specified target. However, the space of possible proteins is too large to
search exhaustively in the laboratory, and functional proteins are scarce in
the vast sequence space. Machine learning (ML) approaches can accelerate
directed evolution by learning to map protein sequences to functions without
building a detailed model of the underlying physics, chemistry and biological
pathways. Despite the great potentials held by these ML methods, they encounter
severe challenges in identifying the most suitable sequences for a targeted
function. These failures can be attributed to the common practice of adopting a
high-dimensional feature representation for protein sequences and inefficient
search methods. To address these issues, we propose an efficient, experimental
design-oriented closed-loop optimization framework for protein directed
evolution, termed ODBO, which employs a combination of novel low-dimensional
protein encoding strategy and Bayesian optimization enhanced with search space
prescreening via outlier detection. We further design an initial sample
selection strategy to minimize the number of experimental samples for training
ML models. We conduct and report four protein directed evolution experiments
that substantiate the capability of the proposed framework for finding of the
variants with properties of interest. We expect the ODBO framework to greatly
reduce the experimental cost and time cost of directed evolution, and can be
further generalized as a powerful tool for adaptive experimental design in a
broader context. |
4,584 | Closing the gap: Exact maximum likelihood training of generative autoencoders using invertible layers | In this work, we provide an exact likelihood alternative to the variational
training of generative autoencoders. We show that VAE-style autoencoders can be
constructed using invertible layers, which offer a tractable exact likelihood
without the need for any regularization terms. This is achieved while leaving
complete freedom in the choice of encoder, decoder and prior architectures,
making our approach a drop-in replacement for the training of existing VAEs and
VAE-style models. We refer to the resulting models as Autoencoders within Flows
(AEF), since the encoder, decoder and prior are defined as individual layers of
an overall invertible architecture. We show that the approach results in
strikingly higher performance than architecturally equivalent VAEs in term of
log-likelihood, sample quality and denoising performance. In a broad sense, the
main ambition of this work is to close the gap between the normalizing flow and
autoencoder literature under the common framework of invertibility and exact
maximum likelihood. |
4,585 | Parallel bandit architecture based on laser chaos for reinforcement learning | Accelerating artificial intelligence by photonics is an active field of study
aiming to exploit the unique properties of photons. Reinforcement learning is
an important branch of machine learning, and photonic decision-making
principles have been demonstrated with respect to the multi-armed bandit
problems. However, reinforcement learning could involve a massive number of
states, unlike previously demonstrated bandit problems where the number of
states is only one. Q-learning is a well-known approach in reinforcement
learning that can deal with many states. The architecture of Q-learning,
however, does not fit well photonic implementations due to its separation of
update rule and the action selection. In this study, we organize a new
architecture for multi-state reinforcement learning as a parallel array of
bandit problems in order to benefit from photonic decision-makers, which we
call parallel bandit architecture for reinforcement learning or PBRL in short.
Taking a cart-pole balancing problem as an instance, we demonstrate that PBRL
adapts to the environment in fewer time steps than Q-learning. Furthermore,
PBRL yields faster adaptation when operated with a chaotic laser time series
than the case with uniformly distributed pseudorandom numbers where the
autocorrelation inherent in the laser chaos provides a positive effect. We also
find that the variety of states that the system undergoes during the learning
phase exhibits completely different properties between PBRL and Q-learning. The
insights obtained through the present study are also beneficial for existing
computing platforms, not just photonic realizations, in accelerating
performances by the PBRL algorithms and correlated random sequences. |
4,586 | Data-driven prediction of Air Traffic Controllers reactions to resolving conflicts | With the aim to enhance automation in conflict detection and resolution
(CD&R) tasks in the Air Traffic Management domain, in this paper we propose
deep learning techniques (DL) that can learn models of Air Traffic Controllers'
(ATCO) reactions in resolving conflicts that can violate separation minimum
constraints among aircraft trajectories: This implies learning when the ATCO
will react towards resolving a conflict, and how he/she will react. Timely
reactions, to which this paper aims, focus on when do reactions happen, aiming
to predict the trajectory points, as the trajectory evolves, that the ATCO
issues a conflict resolution action, while also predicting the type of
resolution action (if any). Towards this goal, the paper formulates the ATCO
reactions prediction problem for CD&R, and presents DL methods that can model
ATCO timely reactions and evaluates these methods in real-world data sets,
showing their efficacy in prediction with very high accuracy. |
4,587 | IFTT-PIN: A PIN-Entry Method Leveraging the Self-Calibration Paradigm | IFTT-PIN is a self-calibrating version of the PIN-entry method introduced in
Roth et al. (2004) [1]. In [1], digits are split into two sets and assigned a
color respectively. To communicate their digit, users press the button with the
same color that is assigned to their digit, which can thus be identified by
elimination after a few iterations. IFTT-PIN uses the same principle but does
not pre-assign colors to each button. Instead, users are free to choose which
button to use for each color. The button-to-color mapping only exists in the
user's mind and is never directly communicated to the interface. In other
words, IFTT-PIN infers both the user's PIN and their preferred button-to-color
mapping at the same time, a process called self-calibration. In this paper, we
present online interactive demonstrations of IFTT-PIN with and without
self-calibration and introduce the key concepts and assumptions making
self-calibration possible. IFTT-PIN can be tested at
https://jgrizou.github.io/IFTT-PIN/ with a video introduction available at
https://youtu.be/5I1ibPJdLHM. We review related work in the field of
brain-computer interface and further propose self-calibration as a novel
approach to protect users against shoulder surfing attacks. Finally, we
introduce a vault cracking challenge as a test of usability and security that
was informally tested at our institute. With IFTT-PIN, we wish to demonstrate a
new interactive experience where users can decide actively and on-the-fly how
to use an interface. The self-calibration paradigm might lead to novel
opportunities for interaction in other applications or domains. We hope this
work will inspire the community to invent them. |
4,588 | Simple Regularisation for Uncertainty-Aware Knowledge Distillation | Considering uncertainty estimation of modern neural networks (NNs) is one of
the most important steps towards deploying machine learning systems to
meaningful real-world applications such as in medicine, finance or autonomous
systems. At the moment, ensembles of different NNs constitute the
state-of-the-art in both accuracy and uncertainty estimation in different
tasks. However, ensembles of NNs are unpractical under real-world constraints,
since their computation and memory consumption scale linearly with the size of
the ensemble, which increase their latency and deployment cost. In this work,
we examine a simple regularisation approach for distribution-free knowledge
distillation of ensemble of machine learning models into a single NN. The aim
of the regularisation is to preserve the diversity, accuracy and uncertainty
estimation characteristics of the original ensemble without any intricacies,
such as fine-tuning. We demonstrate the generality of the approach on
combinations of toy data, SVHN/CIFAR-10, simple to complex NN architectures and
different tasks. |
4,589 | scICML: Information-theoretic Co-clustering-based Multi-view Learning for the Integrative Analysis of Single-cell Multi-omics data | Modern high-throughput sequencing technologies have enabled us to profile
multiple molecular modalities from the same single cell, providing
unprecedented opportunities to assay celluar heterogeneity from multiple
biological layers. However, the datasets generated from these technologies tend
to have high level of noise and are highly sparse, bringing challenges to data
analysis. In this paper, we develop a novel information-theoretic
co-clustering-based multi-view learning (scICML) method for multi-omics
single-cell data integration. scICML utilizes co-clusterings to aggregate
similar features for each view of data and uncover the common clustering
pattern for cells. In addition, scICML automatically matches the clusters of
the linked features across different data types for considering the biological
dependency structure across different types of genomic features. Our
experiments on four real-world datasets demonstrate that scICML improves the
overall clustering performance and provides biological insights into the data
analysis of peripheral blood mononuclear cells. |
4,590 | Variational Inference for Bayesian Bridge Regression | We study the implementation of Automatic Differentiation Variational
inference (ADVI) for Bayesian inference on regression models with bridge
penalization. The bridge approach uses $\ell_{\alpha}$ norm, with $\alpha \in
(0, +\infty)$ to define a penalization on large values of the regression
coefficients, which includes the Lasso ($\alpha = 1$) and ridge $(\alpha = 2)$
penalizations as special cases. Full Bayesian inference seamlessly provides
joint uncertainty estimates for all model parameters. Although MCMC aproaches
are available for bridge regression, it can be slow for large dataset,
specially in high dimensions. The ADVI implementation allows the use of small
batches of data at each iteration (due to stochastic gradient based
algorithms), therefore speeding up computational time in comparison with MCMC.
We illustrate the approach on non-parametric regression models with B-splines,
although the method works seamlessly for other choices of basis functions. A
simulation study shows the main properties of the proposed method. |
4,591 | The Impact of COVID-19 Pandemic on LGBTQ Online Communities | The COVID-19 pandemic has disproportionately impacted the lives of
minorities, such as members of the LGBTQ community (lesbian, gay, bisexual,
transgender, and queer) due to pre-existing social disadvantages and health
disparities. Although extensive research has been carried out on the impact of
the COVID-19 pandemic on different aspects of the general population's lives,
few studies are focused on the LGBTQ population. In this paper, we identify a
group of Twitter users who self-disclose to belong to the LGBTQ community. We
develop and evaluate two sets of machine learning classifiers using a
pre-pandemic and a during pandemic dataset to identify Twitter posts exhibiting
minority stress, which is a unique pressure faced by the members of the LGBTQ
population due to their sexual and gender identities. For this task, we collect
a set of 20,593,823 posts by 7,241 self-disclosed LGBTQ users and annotate a
randomly selected subset of 2800 posts. We demonstrate that our best
pre-pandemic and during pandemic models show strong and stable performance for
detecting posts that contain minority stress. We investigate the linguistic
differences in minority stress posts across pre- and during-pandemic periods.
We find that anger words are strongly associated with minority stress during
the COVID-19 pandemic. We explore the impact of the pandemic on the emotional
states of the LGBTQ population by conducting controlled comparisons with the
general population. We adopt propensity score-based matching to perform a
causal analysis. The results show that the LBGTQ population have a greater
increase in the usage of cognitive words and worsened observable attribute in
the usage of positive emotion words than the group of the general population
with similar pre-pandemic behavioral attributes. |
4,592 | Differentially private Riemannian optimization | In this paper, we study the differentially private empirical risk
minimization problem where the parameter is constrained to a Riemannian
manifold. We introduce a framework of differentially private Riemannian
optimization by adding noise to the Riemannian gradient on the tangent space.
The noise follows a Gaussian distribution intrinsically defined with respect to
the Riemannian metric. We adapt the Gaussian mechanism from the Euclidean space
to the tangent space compatible to such generalized Gaussian distribution. We
show that this strategy presents a simple analysis as compared to directly
adding noise on the manifold. We further show privacy guarantees of the
proposed differentially private Riemannian (stochastic) gradient descent using
an extension of the moments accountant technique. Additionally, we prove
utility guarantees under geodesic (strongly) convex, general nonconvex
objectives as well as under the Riemannian Polyak-{\L}ojasiewicz condition. We
show the efficacy of the proposed framework in several applications. |
4,593 | Spatial Autoregressive Coding for Graph Neural Recommendation | Graph embedding methods including traditional shallow models and deep Graph
Neural Networks (GNNs) have led to promising applications in recommendation.
Nevertheless, shallow models especially random-walk-based algorithms fail to
adequately exploit neighbor proximity in sampled subgraphs or sequences due to
their optimization paradigm. GNN-based algorithms suffer from the insufficient
utilization of high-order information and easily cause over-smoothing problems
when stacking too much layers, which may deteriorate the recommendations of
low-degree (long-tail) items, limiting the expressiveness and scalability. In
this paper, we propose a novel framework SAC, namely Spatial Autoregressive
Coding, to solve the above problems in a unified way. To adequately leverage
neighbor proximity and high-order information, we design a novel spatial
autoregressive paradigm. Specifically, we first randomly mask multi-hop
neighbors and embed the target node by integrating all other surrounding
neighbors with an explicit multi-hop attention. Then we reinforce the model to
learn a neighbor-predictive coding for the target node by contrasting the
coding and the masked neighbors' embedding, equipped with a new hard negative
sampling strategy. To learn the minimal sufficient representation for the
target-to-neighbor prediction task and remove the redundancy of neighbors, we
devise Neighbor Information Bottleneck by maximizing the mutual information
between target predictive coding and the masked neighbors' embedding, and
simultaneously constraining those between the coding and surrounding neighbors'
embedding. Experimental results on both public recommendation datasets and a
real scenario web-scale dataset Douyin-Friend-Recommendation demonstrate the
superiority of SAC compared with state-of-the-art methods. |
4,594 | A Boosting Algorithm for Positive-Unlabeled Learning | Positive-unlabeled (PU) learning deals with binary classification problems
when only positive (P) and unlabeled (U) data are available. A lot of PU
methods based on linear models and neural networks have been proposed; however,
there still lacks study on how the theoretically sound boosting-style
algorithms could work with P and U data. Considering that in some scenarios
when neural networks cannot perform as good as boosting algorithms even with
fully-supervised data, we propose a novel boosting algorithm for PU learning:
Ada-PU, which compares against neural networks. Ada-PU follows the general
procedure of AdaBoost while two different distributions of P data are
maintained and updated. After a weak classifier is learned on the newly updated
distribution, the corresponding combining weight for the final ensemble is
estimated using only PU data. We demonstrated that with a smaller set of base
classifiers, the proposed method is guaranteed to keep the theoretical
properties of boosting algorithms. In experiments, we showed that Ada-PU
outperforms neural networks on benchmark PU datasets. We also study a
real-world dataset UNSW-NB15 in cyber security and demonstrated that Ada-PU has
superior performance for malicious activity detection. |
4,595 | Neural ODEs with Irregular and Noisy Data | Measurement noise is an integral part while collecting data of a physical
process. Thus, noise removal is necessary to draw conclusions from these data,
and it often becomes essential to construct dynamical models using these data.
We discuss a methodology to learn differential equation(s) using noisy and
irregular sampled measurements. In our methodology, the main innovation can be
seen in the integration of deep neural networks with the neural ordinary
differential equations (ODEs) approach. Precisely, we aim at learning a neural
network that provides (approximately) an implicit representation of the data
and an additional neural network that models the vector fields of the dependent
variables. We combine these two networks by constraining using neural ODEs. The
proposed framework to learn a model describing the vector field is highly
effective under noisy measurements. The approach can handle scenarios where
dependent variables are not available at the same temporal grid. Moreover, a
particular structure, e.g., second-order with respect to time, can easily be
incorporated. We demonstrate the effectiveness of the proposed method for
learning models using data obtained from various differential equations and
present a comparison with the neural ODE method that does not make any special
treatment to noise. |
4,596 | Nebula-I: A General Framework for Collaboratively Training Deep Learning Models on Low-Bandwidth Cloud Clusters | The ever-growing model size and scale of compute have attracted increasing
interests in training deep learning models over multiple nodes. However, when
it comes to training on cloud clusters, especially across remote clusters, huge
challenges are faced. In this work, we introduce a general framework, Nebula-I,
for collaboratively training deep learning models over remote heterogeneous
clusters, the connections between which are low-bandwidth wide area networks
(WANs). We took natural language processing (NLP) as an example to show how
Nebula-I works in different training phases that include: a) pre-training a
multilingual language model using two remote clusters; and b) fine-tuning a
machine translation model using knowledge distilled from pre-trained models,
which run through the most popular paradigm of recent deep learning. To balance
the accuracy and communication efficiency, in Nebula-I, parameter-efficient
training strategies, hybrid parallel computing methods and adaptive
communication acceleration techniques are jointly applied. Meanwhile, security
strategies are employed to guarantee the safety, reliability and privacy in
intra-cluster computation and inter-cluster communication. Nebula-I is
implemented with the PaddlePaddle deep learning framework, which can support
collaborative training over heterogeneous hardware, e.g. GPU and NPU.
Experiments demonstrate that the proposed framework could substantially
maximize the training efficiency while preserving satisfactory NLP performance.
By using Nebula-I, users can run large-scale training tasks over cloud clusters
with minimum developments, and the utility of existed large pre-trained models
could be further promoted. We also introduced new state-of-the-art results on
cross-lingual natural language inference tasks, which are generated based upon
a novel learning framework and Nebula-I. |
4,597 | Personalized Interventions for Online Moderation | Current online moderation follows a one-size-fits-all approach, where each
intervention is applied in the same way to all users. This naive approach is
challenged by established socio-behavioral theories and by recent empirical
results that showed the limited effectiveness of such interventions. We propose
a paradigm-shift in online moderation by moving towards a personalized and
user-centered approach. Our multidisciplinary vision combines state-of-the-art
theories and practices in diverse fields such as computer science, sociology
and psychology, to design personalized moderation interventions (PMIs). In
outlining the path leading to the next-generation of moderation interventions,
we also discuss the most prominent challenges introduced by such a disruptive
change. |
4,598 | Why only Micro-F1? Class Weighting of Measures for Relation Classification | Relation classification models are conventionally evaluated using only a
single measure, e.g., micro-F1, macro-F1 or AUC. In this work, we analyze
weighting schemes, such as micro and macro, for imbalanced datasets. We
introduce a framework for weighting schemes, where existing schemes are
extremes, and two new intermediate schemes. We show that reporting results of
different weighting schemes better highlights strengths and weaknesses of a
model. |
4,599 | Neural Network Architecture Beyond Width and Depth | This paper proposes a new neural network architecture by introducing an
additional dimension called height beyond width and depth. Neural network
architectures with height, width, and depth as hyperparameters are called
three-dimensional architectures. It is shown that neural networks with
three-dimensional architectures are significantly more expressive than the ones
with two-dimensional architectures (those with only width and depth as
hyperparameters), e.g., standard fully connected networks. The new network
architecture is constructed recursively via a nested structure, and hence we
call a network with the new architecture nested network (NestNet). A NestNet of
height $s$ is built with each hidden neuron activated by a NestNet of height
$\le s-1$. When $s=1$, a NestNet degenerates to a standard network with a
two-dimensional architecture. It is proved by construction that height-$s$ ReLU
NestNets with $\mathcal{O}(n)$ parameters can approximate Lipschitz continuous
functions on $[0,1]^d$ with an error $\mathcal{O}(n^{-(s+1)/d})$, while the
optimal approximation error of standard ReLU networks with $\mathcal{O}(n)$
parameters is $\mathcal{O}(n^{-2/d})$. Furthermore, such a result is extended
to generic continuous functions on $[0,1]^d$ with the approximation error
characterized by the modulus of continuity. Finally, a numerical example is
provided to explore the advantages of the super approximation power of ReLU
NestNets. |