Datasets:

sswt

/

arxiv_sample_50K

like 0

Asymptotic analysis on charging dynamics for stackelectrode model of supercapacitors ; Supercapacitors are promising electrochemical energy storage devices due to their prominent performance in rapid chargingdischarging rates, long cycle life, stability, etc. Experimental measurement and theoretical prediction on charging timescale for supercapacitors often have large difference. This work develops a matched asymptotic expansion method to derive the charging dynamics of supercapacitors with porous electrodes, in which the supercapacitors are described by the stackelectrode model. Coupling leadingorder solutions between every two stacks by continuity of ionic concentration and fluxes leads to an ODE system, which is a generalized equivalent circuit model for zeta potentials, with the potentialdependent nonlinear capacitance and resistance determined by physical parameters of electrolytes, e.g., specific counterion valences for asymmetric electrolytes. Linearized stability analysis on the ODE system after projection is developed to theoretically characterize the charging timescale. The derived asymptotic solutions are numerically verified. Further numerical investigations on the biexponential charging timescales demonstrate that the proposed generalized equivalent circuit model, as well as companion linearized stability analysis, can faithfully capture the charging dynamics of symmetricasymmetric electrolytes in supercapacitors with porous electrodes.

DiffusionInst Diffusion Model for Instance Segmentation ; Diffusion frameworks have achieved comparable performance with previous stateoftheart image generation models. Researchers are curious about its variants in discriminative tasks because of its powerful noisetoimage denoising pipeline. This paper proposes DiffusionInst, a novel framework that represents instances as instanceaware filters and formulates instance segmentation as a noisetofilter denoising process. The model is trained to reverse the noisy groundtruth without any inductive bias from RPN. During inference, it takes a randomly generated filter as input and outputs mask in onestep or multistep denoising. Extensive experimental results on COCO and LVIS show that DiffusionInst achieves competitive performance compared to existing instance segmentation models with various backbones, such as ResNet and Swin Transformers. We hope our work could serve as a strong baseline, which could inspire designing more efficient diffusion frameworks for challenging discriminative tasks. Our code is available in httpsgithub.comchenhaoxingDiffusionInst.

DialogCC LargeScale MultiModal Dialogue Dataset ; As sharing images in an instant message is a crucial factor, there has been active research on learning a imagetext multimodal dialogue model. However, training a wellgeneralized multimodal dialogue model is challenging because existing multimodal dialogue datasets contain a small number of data, limited topics, and a restricted variety of images per dialogue. In this paper, we present a multimodal dialogue dataset creation pipeline that involves matching largescale images to dialogues based on CLIP similarity. Using this automatic pipeline, we propose a largescale multimodal dialogue dataset, DialogCC, which covers diverse realworld topics and various images per dialogue. With extensive experiments, we demonstrate that training a multimodal dialogue model with our dataset can improve generalization performance. Additionally, existing models trained with our dataset achieve stateoftheart performance on image and text retrieval tasks. The source code and the dataset will be released after publication.

Learning Video Representations from Large Language Models ; We introduce LaViLa, a new approach to learning videolanguage representations by leveraging Large Language Models LLMs. We repurpose pretrained LLMs to be conditioned on visual input, and finetune them to create automatic video narrators. Our autogenerated narrations offer a number of advantages, including dense coverage of long videos, better temporal synchronization of the visual information and text, and much higher diversity of text. The videotext embedding learned contrastively with these additional autogenerated narrations outperforms the previous stateoftheart on multiple firstperson and thirdperson video tasks, both in zeroshot and finetuned setups. Most notably, LaViLa obtains an absolute gain of 10.1 on EGTEA classification and 5.9 EpicKitchens100 multiinstance retrieval benchmarks. Furthermore, LaViLa trained with only half the narrations from the Ego4D dataset outperforms baseline models trained on the full set, and shows positive scaling behavior on increasing pretraining data and model size.

Dynamic TestTime Augmentation via Differentiable Functions ; Distribution shifts, which often occur in the real world, degrade the accuracy of deep learning systems, and thus improving robustness is essential for practical applications. To improve robustness, we study an image enhancement method that generates recognitionfriendly images without retraining the recognition model. We propose a novel image enhancement method, DynTTA, which is based on differentiable data augmentation techniques and generates a blended image from many augmented images to improve the recognition accuracy under distribution shifts. In addition to standard data augmentations, DynTTA also incorporates deep neural networkbased image transformation, which further improves the robustness. Because DynTTA is composed of differentiable functions, it is directly trained with the classification loss of the recognition model. We experiment with widely used image recognition datasets using various classification models, including Vision Transformer and MLPMixer. DynTTA improves the robustness with almost no reduction in classification accuracy for clean images, which is a better result than the existing methods. Furthermore, we show that estimating the training time augmentation for distributionshifted datasets using DynTTA and retraining the recognition model with the estimated augmentation significantly improves robustness.

TRBLLmaker Transformer Reads Between Lyrics Lines maker ; Even for us, it can be challenging to comprehend the meaning of songs. As part of this project, we explore the process of generating the meaning of songs. Despite the widespread use of texttotext models, few attempts have been made to achieve a similar objective. Songs are primarily studied in the context of sentiment analysis. This involves identifying opinions and emotions in texts, evaluating them as positive or negative, and utilizing these evaluations to make music recommendations. In this paper, we present a generative model that offers implicit meanings for several lines of a song. Our model uses a decoder Transformer architecture GPT2, where the input is the lyrics of a song. Furthermore, we compared the performance of this architecture with that of the encoderdecoder Transformer architecture of the T5 model. We also examined the effect of different prompt types with the option of appending additional information, such as the name of the artist and the title of the song. Moreover, we tested different decoding methods with different training parameters and evaluated our results using ROUGE. In order to build our dataset, we utilized the 'Genious' API, which allowed us to acquire the lyrics of songs and their explanations, as well as their rich metadata.

Chai's conjecture for semiabelian Jacobians ; We prove Chai's conjecture on the additivity of the base change conductor of semiabelian varieties in the case of Jacobians of proper curves. This includes the first infinite family of nontrivial wildly ramified examples. Along the way, we extend Raynaud's construction of the N'eron lftmodel of a Jacobian in terms of the Picard functor to arbitrary seminormal curves beyond which Jacobians admit no N'eron lftmodels, as shown by our more general structural results. Finally, we investigate the structure of Jacobians of not necessarily geometrically reduced proper curves over fields of degree of imperfection at most one and prove two conjectures about the existence of N'eron models and N'eron lftmodels due to BoschLutkebohmertRaynaud for Jacobians of general proper curves in the case of perfect residue fields, thus strengthening the author's previous results in this situation.

A Unified Knowledge Graph Augmentation Service for Boosting Domainspecific NLP Tasks ; By focusing the pretraining process on domainspecific corpora, some domainspecific pretrained language models PLMs have achieved stateoftheart results. However, it is underinvestigated to design a unified paradigm to inject domain knowledge in the PLM finetuning stage. We propose KnowledgeDA, a unified domain language model development service to enhance the taskspecific training procedure with domain knowledge graphs. Given domainspecific task texts input, KnowledgeDA can automatically generate a domainspecific language model following three steps i localize domain knowledge entities in texts via an embeddingsimilarity approach; ii generate augmented samples by retrieving replaceable domain entity pairs from two views of both knowledge graph and training data; iii select highquality augmented samples for finetuning via confidencebased assessment. We implement a prototype of KnowledgeDA to learn language models for two domains, healthcare and software development. Experiments on domainspecific text classification and QA tasks verify the effectiveness and generalizability of KnowledgeDA.

CrossModal SimilarityBased Curriculum Learning for Image Captioning ; Image captioning models require the highlevel generalization ability to describe the contents of various images in words. Most existing approaches treat the imagecaption pairs equally in their training without considering the differences in their learning difficulties. Several image captioning approaches introduce curriculum learning methods that present training data with increasing levels of difficulty. However, their difficulty measurements are either based on domainspecific features or prior model training. In this paper, we propose a simple yet efficient difficulty measurement for image captioning using crossmodal similarity calculated by a pretrained visionlanguage model. Experiments on the COCO and Flickr30k datasets show that our proposed approach achieves superior performance and competitive convergence speed to baselines without requiring heuristics or incurring additional training costs. Moreover, the higher model performance on difficult examples and unseen data also demonstrates the generalization ability.

Calibrating AI Models for Wireless Communications via Conformal Prediction ; When used in complex engineered systems, such as communication networks, artificial intelligence AI models should be not only as accurate as possible, but also well calibrated. A wellcalibrated AI model is one that can reliably quantify the uncertainty of its decisions, assigning high confidence levels to decisions that are likely to be correct and low confidence levels to decisions that are likely to be erroneous. This paper investigates the application of conformal prediction as a general framework to obtain AI models that produce decisions with formal calibration guarantees. Conformal prediction transforms probabilistic predictors into set predictors that are guaranteed to contain the correct answer with a probability chosen by the designer. Such formal calibration guarantees hold irrespective of the true, unknown, distribution underlying the generation of the variables of interest, and can be defined in terms of ensemble or timeaveraged probabilities. In this paper, conformal prediction is applied for the first time to the design of AI for communication systems in conjunction to both frequentist and Bayesian learning, focusing on demodulation, modulation classification, and channel prediction.

Improving Chess Commentaries by Combining Language Models with Symbolic Reasoning Engines ; Despite many recent advancements in language modeling, stateoftheart language models lack grounding in the real world and struggle with tasks involving complex reasoning. Meanwhile, advances in the symbolic reasoning capabilities of AI have led to systems that outperform humans in games like chess and Go Silver et al., 2018. Chess commentary provides an interesting domain for bridging these two fields of research, as it requires reasoning over a complex board state and providing analyses in natural language. In this work we demonstrate how to combine symbolic reasoning engines with controllable language models to generate chess commentaries. We conduct experiments to demonstrate that our approach generates commentaries that are preferred by human judges over previous baselines.

SelfPrompting Large Language Models for ZeroShot OpenDomain QA ; OpenDomain Question Answering ODQA aims at answering factoid questions without explicitly providing specific background documents. In a zeroshot setting, this task is more challenging since no data is available to train customized models like RetrieverReaders. Recently, Large Language Models LLMs like GPT3 have shown their power in zeroshot ODQA with direct prompting methods, but these methods are still far from releasing the full powerfulness of LLMs only in an implicitly invoking way. In this paper, we propose a SelfPrompting framework to explicitly utilize the massive knowledge stored in the parameters of LLMs and their strong instruction understanding abilities. Concretely, we prompt LLMs step by step to generate multiple pseudo QA pairs with background passages and explanations from scratch and then use those generated elements for incontext learning. Experimental results show our method surpasses previous SOTA methods significantly on three widelyused ODQA datasets, and even achieves comparable performance with some RetrieverReader models finetuned on full training data.

Inference with approximate local false discovery rates ; Efron's twogroup model is widely used in large scale multiple testing. This model assumes that test statistics are mutually independent, however in realistic settings they are typically dependent, and taking the dependence into account can boost power. The general twogroup model takes the dependence between the test statistics into account. Optimal policies in the general twogroup model require calculation, for each hypothesis, of the probability that it is a true null given all test statistics, denoted local false discovery rate locFDR. Unfortunately, calculating locFDRs under realistic dependence structures can be computationally prohibitive. We propose calculating approximate locFDRs based on a properly defined Nneighborhood for each hypothesis. We prove that by thresholding the approximate locFDRs with a fixed threshold, the marginal false discovery rate is controlled for any dependence structure. Furthermore, we prove that this is the optimal procedure in a restricted class of decision rules, where decision for each hypothesis is only guided by its Nneighborhood. We show through extensive simulations that our proposed method achieves substantial power gains compared to alternative practical approaches, while maintaining conceptual simplicity and computational feasibility. We demonstrate the utility of our method on a genome wide association study of height.

SrTR Selfreasoning Transformer with Visuallinguistic Knowledge for Scene Graph Generation ; Objects in a scene are not always related. The execution efficiency of the onestage scene graph generation approaches are quite high, which infer the effective relation between entity pairs using sparse proposal sets and a few queries. However, they only focus on the relation between subject and object in triplet set subject entity, predicate entity, object entity, ignoring the relation between subject and predicate or predicate and object, and the model lacks selfreasoning ability. In addition, linguistic modality has been neglected in the onestage method. It is necessary to mine linguistic modality knowledge to improve model reasoning ability. To address the abovementioned shortcomings, a Selfreasoning Transformer with Visuallinguistic Knowledge SrTR is proposed to add flexible selfreasoning ability to the model. An encoderdecoder architecture is adopted in SrTR, and a selfreasoning decoder is developed to complete three inferences of the triplet set, sop, spo and pos. Inspired by the largescale pretraining imagetext foundation models, visuallinguistic prior knowledge is introduced and a visuallinguistic alignment strategy is designed to project visual representations into semantic spaces with prior knowledge to aid relational reasoning. Experiments on the Visual Genome dataset demonstrate the superiority and fast inference ability of the proposed method.

Difformer Empowering Diffusion Models on the Embedding Space for Text Generation ; Diffusion models have achieved stateoftheart synthesis quality on both visual and audio tasks, and recent works further adapt them to textual data by diffusing on the embedding space. In this paper, we conduct systematic studies and analyze the challenges between the continuous data space and the embedding space which have not been carefully explored. Firstly, the data distribution is learnable for embeddings, which may lead to the collapse of the loss function. Secondly, as the norm of embeddings varies between popular and rare words, adding the same noise scale will lead to suboptimal results. In addition, we find the normal level of noise causes insufficient training of the model. To address the above challenges, we propose Difformer, an embedding diffusion model based on Transformer, which consists of three essential modules including an additional anchor loss function, a layer normalization module for embeddings, and a noise factor to the Gaussian noise. Experiments on two seminal text generation tasks including machine translation and text summarization show the superiority of Difformer over compared embedding diffusion baselines.

Robust simulation design for generalized linear models in conditions of heteroscedasticity or correlation ; A metamodel of the inputoutput data of a computationally expensive simulation is often employed for prediction, optimization, or sensitivity analysis purposes. Fitting is enabled by a designed experiment, and for computationally expensive simulations, the design efficiency is of importance. Heteroscedasticity in simulation output is common, and it is potentially beneficial to induce dependence through the reuse of pseudorandom number streams to reduce the variance of the metamodel parameter estimators. In this paper, we develop a computational approach to robust design for computer experiments without the need to assume independence or identical distribution of errors. Through explicit inclusion of the variance or correlation structures into the metamodel distribution, either maximum likelihood estimation or generalized estimating equations can be employed to obtain an appropriate Fisher information matrix. Robust designs can then be computationally sought which maximize some relevant summary measure of this matrix, averaged across a prior distribution of any unknown parameters.

To Adapt or to Annotate Challenges and Interventions for Domain Adaptation in OpenDomain Question Answering ; Recent advances in opendomain question answering ODQA have demonstrated impressive accuracy on standard Wikipedia style benchmarks. However, it is less clear how robust these models are and how well they perform when applied to realworld applications in drastically different domains. While there has been some work investigating how well ODQA models perform when tested for outofdomain OOD generalization, these studies have been conducted only under conservative shifts in data distribution and typically focus on a single component ie. retrieval rather than an endtoend system. In response, we propose a more realistic and challenging domain shift evaluation setting and, through extensive experiments, study endtoend model performance. We find that not only do models fail to generalize, but high retrieval scores often still yield poor answer prediction accuracy. We then categorize different types of shifts and propose techniques that, when presented with a new dataset, predict if intervention methods are likely to be successful. Finally, using insights from this analysis, we propose and evaluate several intervention methods which improve endtoend answer F1 score by up to 24 points.

CausalDialogue Modeling Utterancelevel Causality in Conversations ; Despite their widespread adoption, neural conversation models have yet to exhibit natural chat capabilities with humans. In this research, we examine user utterances as causes and generated responses as effects, recognizing that changes in a cause should produce a different effect. To further explore this concept, we have compiled and expanded upon a new dataset called CausalDialogue through crowdsourcing. This dataset includes multiple causeeffect pairs within a directed acyclic graph DAG structure. Our analysis reveals that traditional loss functions struggle to effectively incorporate the DAG structure, leading us to propose a causalityenhanced method called Exponential Maximum Average Treatment Effect ExMATE to enhance the impact of causality at the utterance level in training neural conversation models. To evaluate the needs of considering causality in dialogue generation, we built a comprehensive benchmark on CausalDialogue dataset using different models, inference, and training methods. Through experiments, we find that a causalityinspired loss like ExMATE can improve the diversity and agility of conventional loss function and there is still room for improvement to reach humanlevel quality on this new dataset.

Understanding Stereotypes in Language Models Towards Robust Measurement and ZeroShot Debiasing ; Generated texts from large pretrained language models have been shown to exhibit a variety of harmful, humanlike biases about various demographics. These findings prompted large efforts aiming to understand and measure such effects, with the goal of providing benchmarks that can guide the development of techniques mitigating these stereotypical associations. However, as recent research has pointed out, the current benchmarks lack a robust experimental setup, consequently hindering the inference of meaningful conclusions from their evaluation metrics. In this paper, we extend these arguments and demonstrate that existing techniques and benchmarks aiming to measure stereotypes tend to be inaccurate and consist of a high degree of experimental noise that severely limits the knowledge we can gain from benchmarking language models based on them. Accordingly, we propose a new framework for robustly measuring and quantifying biases exhibited by generative language models. Finally, we use this framework to investigate GPT3's occupational gender bias and propose prompting techniques for mitigating these biases without the need for finetuning.

Extractive Text Summarization Using Generalized Additive Models with Interactions for Sentence Selection ; Automatic Text Summarization ATS is becoming relevant with the growth of textual data; however, with the popularization of public largescale datasets, some recent machine learning approaches have focused on dense models and architectures that, despite producing notable results, usually turn out in models difficult to interpret. Given the challenge behind interpretable learningbased text summarization and the importance it may have for evolving the current state of the ATS field, this work studies the application of two modern Generalized Additive Models with interactions, namely Explainable Boosting Machine and GAMINet, to the extractive summarization problem based on linguistic features and binary classification.

Quantum sunburst model under interaction quench entanglement and role of initial state coherence ; We study the nonequilibrium dynamics of an isolated bipartite quantum system, sunburst model, under interaction quench. The prequench limit of this model is two noninteracting integrable systems, one a transverse Ising chain while the other is finite number of isolated qubits. As a function of interaction strength, the spectral fluctuation properties goes from Poisson to WignerDyson statistics. We chose entanglement entropy as a probe to study approach to thermalization or lack of it in postquench dynamics. In near integrable limit, as expected, the linear entropy displays oscillatory behaviour while in chaotic limit, it saturates. Along with the chaotic nature of the time evolution generator, we show the importance of the role played by coherence of initial state in deciding the nature of thermalization. We further show that these findings are general by replacing Ising ring by disordered XXZ model with disorder strength putting it in manybody localized phase.

Localization and topological transitions in generalized nonHermitian SSH models ; We study the localization and topological transitions of the generalized nonHermitian SSH models, where the nonHermiticities are introduced by the complex quasiperiodic hopping and the nonreciprocal hopping. We elucidate the universality of the models and how many models can be mapped to them. Under the open boundary condition, two delocalization transitions are found due to the competition between the Anderson localization and the boundary localization from the nontrivial edge states and the nonHermitian skin effect. Under the periodic boundary condition, only one delocalization transition is found due to the disappearance of the nonHermitian skin effect. The winding numbers of energy and the Lyapunov exponents in analytical form are obtained to exactly characterize the two deloaclizateon transitions. It finds that the delocalization transitions don't accompany the topological transition. Furthermore, the large onsite nonHermiticity and the large nonreciprocal hopping are all detrimental to the topological transitions. However, the large nonreciprocal hopping enhances the Anderson localizations. The above analyses are verified by calculating the energy gap and the inverse of the participation ratio numerically.

CellTranspose Fewshot Domain Adaptation for Cellular Instance Segmentation ; Automated cellular instance segmentation is a process utilized for accelerating biological research for the past two decades, and recent advancements have produced higher quality results with less effort from the biologist. Most current endeavors focus on completely cutting the researcher out of the picture by generating highly generalized models. However, these models invariably fail when faced with novel data, distributed differently than the ones used for training. Rather than approaching the problem with methods that presume the availability of large amounts of target data and computing power for retraining, in this work we address the even greater challenge of designing an approach that requires minimal amounts of new annotated data as well as training time. We do so by designing specialized contrastive losses that leverage the few annotated samples very efficiently. A large set of results show that 3 to 5 annotations lead to models with accuracy that 1 significantly mitigate the covariate shift effects; 2 matches or surpasses other adaptation methods; 3 even approaches methods that have been fully retrained on the target distribution. The adaptation training is only a few minutes, paving a path towards a balance between model performance, computing requirements and expertlevel annotation needs.

NeRFGaze A HeadEye Redirection Parametric Model for Gaze Estimation ; Gaze estimation is the fundamental basis for many visual tasks. Yet, the high cost of acquiring gaze datasets with 3D annotations hinders the optimization and application of gaze estimation models. In this work, we propose a novel HeadEye redirection parametric model based on Neural Radiance Field, which allows dense gaze data generation with view consistency and accurate gaze direction. Moreover, our headeye redirection parametric model can decouple the face and eyes for separate neural rendering, so it can achieve the purpose of separately controlling the attributes of the face, identity, illumination, and eye gaze direction. Thus diverse 3Daware gaze datasets could be obtained by manipulating the latent code belonging to different face attributions in an unsupervised manner. Extensive experiments on several benchmarks demonstrate the effectiveness of our method in domain generalization and domain adaptation for gaze estimation tasks.

Vacuum Stability and Radiative Symmetry Breaking of the ScaleInvariant Singlet Extension of Type II Seesaw Model ; The questions of the origin of electroweak symmetry breaking and neutrino mass are two major puzzles in particle physics. Neutrino mass generation requires new physics beyond the Standard Model and also suggests reconsideration of physics of symmetry breaking. The aim of this paper is to study radiative symmetry breaking in the singlet scalar extension of type II seesaw neutrino mass model. We derive boundedfrombelow conditions for the scalar potential of the model in full generality for the first time. The GildenerWeinberg approach is utilised in minimising the multiscalar potential. Upon imposing the boundedfrombelow and perturbativity conditions, as well as experimental constraints from colliders, we find the parameter space of scalar quartic couplings that can radiatively realise electroweak symmetry breaking at oneloop level. To satisfy all the constraints, the masses of the heavy tripletlike Higgs bosons must be nearly degenerate. The evolution of the Higgs doublet quartic coupling lambdaH can be prevented from being negative up to the Planck scale.

DataDriven Optimization of Directed Information over Discrete Alphabets ; Directed information DI is a fundamental measure for the study and analysis of sequential stochastic models. In particular, when optimized over input distributions it characterizes the capacity of general communication channels. However, analytic computation of DI is typically intractable and existing optimization techniques over discrete input alphabets require knowledge of the channel model, which renders them inapplicable when only samples are available. To overcome these limitations, we propose a novel estimationoptimization framework for DI over discrete input spaces. We formulate DI optimization as a Markov decision process and leverage reinforcement learning techniques to optimize a deep generative model of the input process probability mass function PMF. Combining this optimizer with the recently developed DI neural estimator, we obtain an endtoend estimationoptimization algorithm which is applied to estimating the feedforward and feedback capacity of various discrete channels with memory. Furthermore, we demonstrate how to use the optimized PMF model to i obtain theoretical bounds on the feedback capacity of unifilar finitestate channels; and ii perform probabilistic shaping of constellations in the peak powerconstrained additive white Gaussian noise channel.

A Residual Diffusion Model for High Perceptual Quality Codec Augmentation ; Diffusion probabilistic models have recently achieved remarkable success in generating high quality image and video data. In this work, we build on this class of generative models and introduce a method for lossy compression of high resolution images. The resulting codec, which we call DIffusonbased Residual Augmentation Codec DIRAC, is the first neural codec to allow smooth traversal of the ratedistortionperception tradeoff at test time, while obtaining competitive performance with GANbased methods in perceptual quality. Furthermore, while sampling from diffusion probabilistic models is notoriously expensive, we show that in the compression setting the number of steps can be drastically reduced.

Resilient Model Predictive Control of Distributed Systems Under Attack Using Local Attack Identification ; With the growing share of renewable energy sources, the uncertainty in power supply is increasing. In addition to the inherent fluctuations in the renewables, this is due to the threat of deliberate malicious attacks, which may become more revalent with a growing number of distributed generation units. Also in other safetycritical technology sectors, control systems are becoming more and more decentralized, causing the targets for attackers and thus the risk of attacks to increase. It is thus essential that distributed controllers are robust toward these uncertainties and able to react quickly to disturbances of any kind. To this end, we present novel methods for modelbased identification of attacks and combine them with distributed model predictive control to obtain a resilient framework for adaptively robust control. The methodology is specially designed for distributed setups with limited local information due to privacy and security reasons. To demonstrate the efficiency of the method, we introduce a mathematical model for physically coupled microgrids under the uncertain influence of renewable generation and adversarial attacks, and perform numerical experiments, applying the proposed method for microgrid control.

Using Large TexttoImage Models with Structured Prompts for Skin Disease Identification A Case Study ; This paper investigates the potential usage of large texttoimage LTI models for the automated diagnosis of a few skin conditions with rarity or a serious lack of annotated datasets. As the input to the LTI model, we provide the targeted instantiation of a generic but succinct prompt structure designed upon careful observations of the conditional narratives from the standard medical textbooks. In this regard, we pave the path to utilizing accessible textbook descriptions for automated diagnosis of conditions with data scarcity through the lens of LTI models. Experiments show the efficacy of the proposed framework, including much better localization of the infected regions. Moreover, it has the immense possibility for generalization across the medical subdomains, not only to mitigate the data scarcity issue but also to debias automated diagnostics from the allpervasive racial biases.

Inflation in fR,T gravity with observational constraints ; The scenario of slowroll inflation is explored in the fR,T theory of gravity where a nonminimal coupling between matter and curvature is included. A noncanonical scalar field is assumed to play the role of inflaton which contains generalized kinetic energy. The study is performed by taking the HamiltonJacobi formalism where the Hubble parameter is taken as a function of the scalar field. In this regard, a powerlaw function and an exponential function of the scalar field are assumed for the Hubble parameter and the model is considered in detail. By performing Python coding and applying the observational data, the free parameters of the model are determined for which the model is put in perfect consistency with the data. Then, using the results, the validity of the swampland criteria and TCC is considered. It is realized that not only the model comes to a good agreement with data, but it also could satisfy the swampland criteria.

Gluon generalized parton distributions and angular momentum in a lightcone spectator model ; We study the leading twist gluon generalized parton distributions GPDs and the gluon angular momentum inside the proton within a lightcone spectator model. Using the lightcone wave functions derived from the model, we provide the expressions of these distributions at the particular kinematical point xi0 in the overlap representation. The numerical results of the Hg, Eg, tildeHg, HTg and ETg as functions of x at different DeltaT are presented. Particularly, Hg, tildeHg at nonzero DeltaT are different from their forward counterparts, the unpolarized distribution f1g and the helicity distribution g1g, respectively. We also obtain the total angular momentum of the gluon contributed to the proton spin Jg0.19, which is consistent with the recent lattice calculation after the uncertainties is considered. The kinetic orbital angular momentum is also calculated and is negative in our model.

Explaining the effects of nonconvergent sampling in the training of EnergyBased Models ; In this paper, we quantify the impact of using nonconvergent Markov chains to train EnergyBased models EBMs. In particular, we show analytically that EBMs trained with nonpersistent short runs to estimate the gradient can perfectly reproduce a set of empirical statistics of the data, not at the level of the equilibrium measure, but through a precise dynamical process. Our results provide a firstprinciples explanation for the observations of recent works proposing the strategy of using short runs starting from random initial conditions as an efficient way to generate highquality samples in EBMs, and lay the groundwork for using EBMs as diffusion models. After explaining this effect in generic EBMs, we analyze two solvable models in which the effect of the nonconvergent sampling in the trained parameters can be described in detail. Finally, we test these predictions numerically on a ConvNet EBM and a Boltzmann machine.

MTTN MultiPair Text to Text Narratives for Prompt Generation ; The increased interest in diffusion models has opened up opportunities for advancements in generative text modeling. These models can produce impressive images when given a wellcrafted prompt, but creating a powerful or meaningful prompt can be hitormiss. To address this, we have created a largescale dataset that is derived and synthesized from real prompts and indexed with popular imagetext datasets such as MSCOCO and Flickr. We have also implemented stages that gradually reduce context and increase complexity, which will further enhance the output due to the complex annotations created. The dataset, called MTTN, includes over 2.4 million sentences divided into 5 stages, resulting in a total of over 12 million pairs, and a vocabulary of over 300,000 unique words, providing ample variation. The original 2.4 million pairs are designed to reflect the way language is used on the internet globally, making the dataset more robust for any model trained on it.

simple diffusion Endtoend diffusion for high resolution images ; Currently, applying diffusion models in pixel space of high resolution images is difficult. Instead, existing approaches focus on diffusion in lower dimensional spaces latent diffusion, or have multiple superresolution levels of generation referred to as cascades. The downside is that these approaches add additional complexity to the diffusion framework. This paper aims to improve denoising diffusion for high resolution images while keeping the model as simple as possible. The paper is centered around the research question How can one train a standard denoising diffusion models on high resolution images, and still obtain performance comparable to these alternate approaches The four main findings are 1 the noise schedule should be adjusted for high resolution images, 2 It is sufficient to scale only a particular part of the architecture, 3 dropout should be added at specific locations in the architecture, and 4 downsampling is an effective strategy to avoid high resolution feature maps. Combining these simple yet effective techniques, we achieve stateoftheart on image generation among diffusion models without sampling modifiers on ImageNet.

QD3SET1 A Database with Quantum Dissipative Dynamics Data Sets ; Simulations of the dynamics of dissipative quantum systems utilize many methods such as physicsbased quantum, semiclassical, and quantumclassical as well as machine learningbased approximations, development and testing of which requires diverse data sets. Here we present a new database QD3SET1 containing eight data sets of quantum dynamical data for two systems of broad interest, spinboson SB model and the FennaMatthewsOlson FMO complex, generated with two different methods solving the dynamics, approximate local thermalizing Lindblad master equation LTLME and highly accurate hierarchy equations of motion HEOM. One data set was generated with the SB model which is a twolevel quantum system coupled to a harmonic environment using HEOM for 1,000 model parameters. Seven data sets were collected for the FMO complex of different sizes7 and 8site monomer and 24site trimer with LTLME and 8site monomer with HEOM for 500879 model parameters. Our QD3SET1 database contains both population and coherence dynamics data and part of it has been already used for machine learningbased quantum dynamics studies.

Gesture Control of Microdrone A LightweightNet with Domain Randomization and Trajectory Generators ; Microdrones can be integrated into various industrial applications but are constrained by their computing power and expert pilots, a secondary challenge. This study presents a computationallyefficient deep convolutional neural network that utilizes Gabor filters and spatial separable convolutions with low computational complexities. An attention module is integrated with the model to complement the performance. Further, perceptionbased action space and trajectory generators are integrated with the model's predictions for intuitive navigation. The computationallyefficient model aids a human operator in controlling a microdrone via gestures. Nearly 18 of computational resources are conserved using the NVIDIA GPU profiler during training. Using a lowcost DJI Tello drone for experiment verification, the computationallyefficient model shows promising results compared to a stateoftheart and conventional computer visionbased technique.

Distilling InternetScale VisionLanguage Models into Embodied Agents ; Instructionfollowing agents must ground language into their observation and action spaces. Learning to ground language is challenging, typically requiring domainspecific engineering or large quantities of human interaction data. To address this challenge, we propose using pretrained visionlanguage models VLMs to supervise embodied agents. We combine ideas from model distillation and hindsight experience replay HER, using a VLM to retroactively generate language describing the agent's behavior. Simple prompting allows us to control the supervision signal, teaching an agent to interact with novel objects based on their names e.g., planes or their features e.g., colors in a 3D rendered environment. Fewshot prompting lets us teach abstract category membership, including preexisting categories food vs toys and adhoc ones arbitrary preferences over objects. Our work outlines a new and effective way to use internetscale VLMs, repurposing the generic language grounding acquired by such models to teach taskrelevant groundings to embodied agents.

Effective Random Test Generation for Deep Learning Compilers ; Deep learning compilers help address difficulties of deploying deep learning models on diverse types of hardware. Testing deep learning compilers is highly crucial, because they are impacting countless AI applications that use them for model optimization and deployment. To test deep learning compilers, random testing, being popularly used for compiler testing practices, faces the challenge of generating semantically valid test inputs, i.e., deep learning models that satisfy the semantic model specifications in short as semantic specifications. To tackle this challenge, in this paper, we propose a novel approach named Isra, including a domainspecific constraint solver that resolves the constraints from the semantic specifications without backtracking. We implement and apply our approach on three popular realworld deep learning compilers including TVM, Glow, and a commercial compiler. The evaluation results show that Isra is more effective than the stateoftheart approaches and the baseline approaches on constructing valid test inputs for compilerbug detection, and Isra successfully finds 24 previously unknown bugs in released versions of the three compilers. These results indicate effectiveness and practical value of Isra.

Deep COVID19 Forecasting for Multiple States with Data Augmentation ; In this work, we propose a deep learning approach to forecasting statelevel COVID19 trends of weekly cumulative death in the United States US and incident cases in Germany. This approach includes a transformer model, an ensemble method, and a data augmentation technique for time series. We arrange the inputs of the transformer in such a way that predictions for different states can attend to the trends of the others. To overcome the issue of scarcity of training data for this COVID19 pandemic, we have developed a novel data augmentation technique to generate useful data for training. More importantly, the generated data can also be used for model validation. As such, it has a twofold advantage 1 more actual observations can be used for training, and 2 the model can be validated on data which has distribution closer to the expected situation. Our model has achieved some of the best statelevel results on the COVID19 Forecast Hub for the US and for Germany.

GANbased Vertical Federated Learning for Label Protection in Binary Classification ; Split learning splitNN has emerged as a popular strategy for addressing the high computational costs and low modeling efficiency in Vertical Federated Learning VFL. However, despite its popularity, vanilla splitNN lacks encryption protection, leaving it vulnerable to privacy leakage issues, especially Label Leakage from Gradients LLG. Motivated by the LLG issue resulting from the use of labels during training, we propose the Generative Adversarial Federated Model GAFM, a novel method designed specifically to enhance label privacy protection by integrating splitNN with Generative Adversarial Networks GANs. GAFM leverages GANs to indirectly utilize label information by learning the label distribution rather than relying on explicit labels, thereby mitigating LLG. GAFM also employs an additional crossentropy loss based on the noisy labels to further improve the prediction accuracy. Our ablation experiment demonstrates that the combination of GAN and the crossentropy loss component is necessary to enable GAFM to mitigate LLG without significantly compromising the model utility. Empirical results on various datasets show that GAFM achieves a better and more robust tradeoff between model utility and privacy compared to all baselines across multiple random runs. In addition, we provide experimental justification to substantiate GAFM's superiority over splitNN, demonstrating that it offers enhanced label protection through gradient perturbation relative to splitNN.

Leveraging Domain Relations for Domain Generalization ; Distribution shift is a major challenge in machine learning, as models often perform poorly during the test stage if the test distribution differs from the training distribution. In this paper, we focus on domain shifts, which occur when the model is applied to new domains that are different from the ones it was trained on, and propose a new approach called D3G. Unlike previous approaches that aim to learn a single model that is domain invariant, D3G learns domainspecific models by leveraging the relations among different domains. Concretely, D3G learns a set of trainingdomainspecific functions during the training stage and reweights them based on domain relations during the test stage. These domain relations can be directly derived or learned from fixed domain metadata. Under mild assumptions, we theoretically proved that using domain relations to reweight trainingdomainspecific functions achieves stronger generalization compared to averaging them. Empirically, we evaluated the effectiveness of D3G using both toy and realworld datasets for tasks such as temperature regression, land use classification, and moleculeprotein interaction prediction. Our results showed that D3G consistently outperformed stateoftheart methods, with an average improvement of 10.6 in performance.

An Informative Path Planning Framework for Active Learning in UAVbased Semantic Mapping ; Unmanned aerial vehicles UAVs are frequently used for aerial mapping and general monitoring tasks. Recent progress in deep learning enabled automated semantic segmentation of imagery to facilitate the interpretation of largescale complex environments. Commonly used supervised deep learning for segmentation relies on large amounts of pixelwise labelled data, which is tedious and costly to annotate. The domainspecific visual appearance of aerial environments often prevents the usage of models pretrained on publicly available datasets. To address this, we propose a novel general planning framework for UAVs to autonomously acquire informative training images for model retraining. We leverage multiple acquisition functions and fuse them into probabilistic terrain maps. Our framework combines the mapped acquisition function information into the UAV's planning objectives. In this way, the UAV adaptively acquires informative aerial images to be manually labelled for model retraining. Experimental results on realworld data and in a photorealistic simulation show that our framework maximises model performance and drastically reduces labelling efforts. Our mapbased planners outperform stateoftheart local planning.

Random Walk on a Rough Surface Renormalization Group Analysis of a Simple Model ; The field theoretic renormalization group is applied to a simple model of random walk on a rough fluctuating surface. We consider the FokkerPlanck equation for a particle in a uniform gravitational field. The surface is modelled by the generalized EdwardsWilkinson linear stochastic equation for the height field. The full stochastic model is reformulated as a multiplicatively renormalizable field theory, which allows for application of the standard renormalization theory. The renormalization group equations have several fixed points that correspond to possible scaling regimes in the infrared range long times, large distances; all the critical dimensions are found exactly. As an example, the spreading law for particle's cloud is derived. It has the form R2tsimeq t2Deltaomega with the exactly known critical dimension of frequency Deltaomega and, in general, differs from the standard expression R2tsimeq t for ordinary random walk.

Concept Algebra for ScoreBased Conditional Models ; This paper concerns the structure of learned representations in textguided generative models, focusing on scorebased models. Here, we focus on the idea that concepts are encoded as subspaces or directions of some representation space. We develop a mathematical formalization of this idea.Using this formalism, we show there's a natural choice of representation with this property, and we develop a simple method for identifying the part of the representation corresponding to a given concept. In particular, this allows us to manipulate the concepts expressed by the model through algebraic manipulation of the representation. We demonstrate the idea with examples textguided image generation, using Stable Diffusion.

Inverse Models for Estimating the Initial Condition of SpatioTemporal AdvectionDiffusion Processes ; Inverse problems involve making inference about unknown parameters of a physical process using observational data. This paper investigates an important class of inverse problems the estimation of the initial condition of a spatiotemporal advectiondiffusion process using spatially sparse data streams. Three spatial sampling schemes are considered, including irregular, nonuniform and shifted uniform sampling. The irregular sampling scheme is the general scenario, while computationally efficient solutions are available in the spectral domain for nonuniform and shifted uniform sampling. For each sampling scheme, the inverse problem is formulated as a regularized convex optimization problem that minimizes the distance between forward model outputs and observations. The optimization problem is solved by the Alternating Direction Method of Multipliers algorithm, which also handles the situation when a linear inequality constraint e.g., nonnegativity is imposed on the model output. Numerical examples are presented, code is made available on GitHub, and discussions are provided to generate some useful insights of the proposed inverse modeling approaches.

CrossModal FineTuning Align then Refine ; Finetuning largescale pretrained models has led to tremendous progress in wellstudied modalities such as vision and NLP. However, similar gains have not been observed in many other modalities due to a lack of relevant pretrained models. In this work, we propose ORCA, a general crossmodal finetuning framework that extends the applicability of a single largescale pretrained model to diverse modalities. ORCA adapts to a target task via an alignthenrefine workflow given the target input, ORCA first learns an embedding network that aligns the embedded feature distribution with the pretraining modality. The pretrained model is then finetuned on the embedded data to exploit the knowledge shared across modalities. Through extensive experiments, we show that ORCA obtains stateoftheart results on 3 benchmarks containing over 60 datasets from 12 modalities, outperforming a wide range of handdesigned, AutoML, generalpurpose, and taskspecific methods. We highlight the importance of data alignment via a series of ablation studies and demonstrate ORCA's utility in datalimited regimes.

Doubly stochastic continuous time random walk ; Since its introduction, some sixty years ago, the MontrollWeiss continuous time random walk has found numerous applications due its ease of use and ability to describe both regular and anomalous diffusion. Yet, despite its broad applicability and generality, the model cannot account for effects coming from random diffusivity fluctuations which have been observed in the motion of asset prices and molecules. To bridge this gap, we introduce a doubly stochastic version of the model in which waiting times between jumps are replaced with a fluctuating jump rate. We show that this newly added layer of randomness gives rise to a rich phenomenology while keeping the model fully tractable allowing us to explore general properties and illustrate them with examples. In particular, we show that the model presented herein provides an alternative pathway to Brownian yet nonGaussian diffusion which has been observed and explained via diffusing diffusivity approaches.

Wireless Communications with SpaceTime Modulated Metasurfaces ; Spacetime modulated metasurfaces STMMs are a newly investigated technology for next 6G generation wireless communication networks. An STMM augments the spatial phase function with a timevarying one across the elements, allowing for the conveyance of information that possibly modulates the impinging signal. Hence, STMM represents an evolution of reconfigurable intelligent surfaces RIS, which only design the spatial phase pattern. STMMs convey signals without a relevant increase in the energy budget, which is convenient for applications where energy is a strong constraint. This paper proposes a mathematical model for STMMbased wireless communication, that creates the basics for two potential STMM architectures. One has excellent design flexibility, whereas the other is more costeffective. The model describes STMM's distinguishing features, such as spacetime coupling, and their impact on system performance. The proposed STMM model addresses the design criteria of a fullduplex system architecture, in which the temporal signal originating at the STMM generates a modulation overlapped with the incident one. The presented numerical results demonstrate the efficacy of the proposed model and its potential to revolutionize wireless communication.

LDFA Latent Diffusion Face Anonymization for Selfdriving Applications ; In order to protect vulnerable road users VRUs, such as pedestrians or cyclists, it is essential that intelligent transportation systems ITS accurately identify them. Therefore, datasets used to train perception models of ITS must contain a significant number of vulnerable road users. However, data protection regulations require that individuals are anonymized in such datasets. In this work, we introduce a novel deep learningbased pipeline for face anonymization in the context of ITS. In contrast to related methods, we do not use generative adversarial networks GANs but build upon recent advances in diffusion models. We propose a twostage method, which contains a face detection model followed by a latent diffusion model to generate realistic face inpaintings. To demonstrate the versatility of anonymized images, we train segmentation methods on anonymized data and evaluate them on nonanonymized data. Our experiment reveal that our pipeline is better suited to anonymize data for segmentation than naive methods and performes comparably with recent GANbased methods. Moreover, face detectors achieve higher mAP scores for faces anonymized by our method compared to naive or recent GANbased methods.

Personalizing Federated Learning with OvertheAir Computations ; Federated edge learning is a promising technology to deploy intelligence at the edge of wireless networks in a privacypreserving manner. Under such a setting, multiple clients collaboratively train a global generic model under the coordination of an edge server. But the training efficiency is often throttled by challenges arising from limited communication and data heterogeneity. This paper presents a distributed training paradigm that employs analog overtheair computation to address the communication bottleneck. Additionally, we leverage a bilevel optimization framework to personalize the federated learning model so as to cope with the data heterogeneity issue. As a result, it enhances the generalization and robustness of each client's local model. We elaborate on the model training procedure and its advantages over conventional frameworks. We provide a convergence analysis that theoretically demonstrates the training efficiency. We also conduct extensive experiments to validate the efficacy of the proposed framework.

Promptbased Learning for Text Readability Assessment ; We propose the novel adaptation of a pretrained seq2seq model for readability assessment. We prove that a seq2seq model T5 or BART can be adapted to discern which text is more difficult from two given texts pairwise. As an exploratory study to promptlearn a neural network for text readability in a texttotext manner, we report useful tips for future work in seq2seq training and rankingbased approach to readability assessment. Specifically, we test nine inputoutput formatsprefixes and show that they can significantly influence the final model performance. Also, we argue that the combination of texttotext training and pairwise ranking setup 1 enables leveraging multiple parallel text simplification data for teaching readability and 2 trains a neural model for the general concept of readability therefore, better crossdomain generalization. At last, we report a 99.6 pairwise classification accuracy on Newsela and a 98.7 for OneStopEnglish, through a joint training approach.

The Second Order Scattering Fading Model with Fluctuating LineofSight ; We present a generalization of the notoriously unwieldy secondorder scattering fading model, which is helpful to alleviate its mathematical complexity while providing an additional degree of freedom. This is accomplished by allowing its dominant specular component associated to lineofsight propagation to randomly fluctuate. The statistical characterization of the newly proposed model is carried out, providing closedform expressions for its probability and cumulative distribution functions, as well as for its generalized Laplacedomain statistics and raw moments. We exemplify how performance analysis can be done in this scenario, and discuss the role of the fading model parameters on system performance.

Line Defect RG Flows in the varepsilon Expansion ; A general analysis of line defect renormalisation group RG flows in the varepsilon expansion below d4 dimensions is undertaken. The defect beta function for general scalarfermion bulk theories is computed to nexttoleading order in the bulk couplings. Scalar models as well as scalarfermion models with various global symmetries in the bulk are considered at leading nontrivial order. Different types of potential infrared IR defect conformal field theories dCFTs and their RG stability are discussed. The possibility of multiple IR stable dCFTs is realised in specific examples with hypertetrahedral symmetry in the bulk. The onepoint function coefficient of the order parameter in the stable IR dCFT of the cubic model is computed at nexttoleading order and compared with that in the IR dCFT of the Heisenberg model.

Mr.Keynes and the... Complexity A suggested agentbased version of the General Theory of Employment, Interest and Money ; This paper presents a model with the aim to follow, as closely as possible, the rationale of the macroeconomic model advanced by J.M. Keynes in his famous The General Theory of Employment, Interest and Money, in order to provide a viable tool for macroeconomic research. Keynes' main result will be shown, i.e., to determine the level of income and employment starting from the marginal efficiency of capital and the marginal propensity to consume, given the interest rate. Elements of the model will be described by referring to the original text. The sequentiality in model operation will prove quintessential in order to describe the complex nature of macroeconomic systems.

Implications of multiaxion dark matter on structure formation ; Axions are candidates for dark matter in the universe.We develop an accurate Boltzmann code to calculate the linear growth of the plasma. As an interesting example, we investigate a mixed dark matter model consisting of cold dark matter CDM and twoaxion dark matter. We analyze the growth of the structure numerically and analytically. We find that an effective single axion with an effective mass and an effective abundance is useful to characterize the twoaxion cosmology. Moreover, we generalize the effective single axion description to multiaxion dark matter cosmology. We also compare the results with those of warm dark matter WDM model. Moreover, we calculate halo mass functions for the mixed model by using the PressSchechter model and linear perturbations and then determine the mass function as a function of masses and axion abundance.

Revisiting Adversarial Training for ImageNet Architectures, Training and Generalization across Threat Models ; While adversarial training has been extensively studied for ResNet architectures and low resolution datasets like CIFAR, much less is known for ImageNet. Given the recent debate about whether transformers are more robust than convnets, we revisit adversarial training on ImageNet comparing ViTs and ConvNeXts. Extensive experiments show that minor changes in architecture, most notably replacing PatchStem with ConvStem, and training scheme have a significant impact on the achieved robustness. These changes not only increase robustness in the seen ellinftythreat model, but even more so improve generalization to unseen ell1ell2robustness. Our modified ConvNeXt, ConvNeXt ConvStem, yields the most robust models across different ranges of model parameters and FLOPs.

Regularized Vector Quantization for Tokenized Image Synthesis ; Quantizing images into discrete representations has been a fundamental problem in unified generative modeling. Predominant approaches learn the discrete representation either in a deterministic manner by selecting the bestmatching token or in a stochastic manner by sampling from a predicted distribution. However, deterministic quantization suffers from severe codebook collapse and misalignment with inference stage while stochastic quantization suffers from low codebook utilization and perturbed reconstruction objective. This paper presents a regularized vector quantization framework that allows to mitigate above issues effectively by applying regularization from two perspectives. The first is a prior distribution regularization which measures the discrepancy between a prior token distribution and the predicted token distribution to avoid codebook collapse and low codebook utilization. The second is a stochastic mask regularization that introduces stochasticity during quantization to strike a good balance between inference stage misalignment and unperturbed reconstruction objective. In addition, we design a probabilistic contrastive loss which serves as a calibrated metric to further mitigate the perturbed reconstruction objective. Extensive experiments show that the proposed quantization framework outperforms prevailing vector quantization methods consistently across different generative models including autoregressive models and diffusion models.

Generating multiplechoice questions for medical question answering with distractors and cuemasking ; Medical multiplechoice question answering MCQA is particularly difficult. Questions may describe patient symptoms and ask for the correct diagnosis, which requires domain knowledge and complex reasoning. Standard language modeling pretraining alone is not sufficient to achieve the best results. citetjin2020disease showed that focusing masked language modeling on disease name prediction when using medical encyclopedic paragraphs as input leads to considerable MCQA accuracy improvement. In this work, we show that 1 finetuning on generated MCQA dataset outperforms the masked language modeling based objective and 2 correctly masking the cues to the answers is critical for good performance. We release new pretraining datasets and achieve stateoftheart results on 4 MCQA datasets, notably 5.7 with basesize model on MedQAUSMLE.

Score Attack A Lower Bound Technique for Optimal Differentially Private Learning ; Achieving optimal statistical performance while ensuring the privacy of personal data is a challenging yet crucial objective in modern data analysis. However, characterizing the optimality, particularly the minimax lower bound, under privacy constraints is technically difficult. To address this issue, we propose a novel approach called the score attack, which provides a lower bound on the differentialprivacyconstrained minimax risk of parameter estimation. The score attack method is based on the tracing attack concept in differential privacy and can be applied to any statistical model with a welldefined score statistic. It can optimally lower bound the minimax risk of estimating unknown model parameters, up to a logarithmic factor, while ensuring differential privacy for a range of statistical problems. We demonstrate the effectiveness and optimality of this general method in various examples, such as the generalized linear model in both classical and highdimensional sparse settings, the BradleyTerryLuce model for pairwise comparisons, and nonparametric regression over the Sobolev class.

Merging Decision Transformers Weight Averaging for Forming MultiTask Policies ; Recent work has shown the promise of creating generalist, transformerbased, models for language, vision, and sequential decisionmaking problems. To create such models, we generally require centralized training objectives, data, and compute. It is of interest if we can more flexibly create generalist policies by merging together multiple, taskspecific, individually trained policies. In this work, we take a preliminary step in this direction through merging, or averaging, subsets of Decision Transformers in parameter space trained on different MuJoCo locomotion problems, forming multitask models without centralized training. We also demonstrate the importance of various methodological choices when merging policies, such as utilizing common pretrained initializations, increasing model capacity, and utilizing Fisher information for weighting parameter importance. In general, we believe research in this direction could help democratize and distribute the process that forms multitask robotics policies. Our implementation is available at httpsgithub.comdaniellawson9999mergingdecisiontransformers.

Sequential threeway decisions with a single hidden layer feedforward neural network ; The threeway decisions strategy has been employed to construct network topology in a single hidden layer feedforward neural network SFNN. However, this model has a general performance, and does not consider the process costs, since it has fixed threshold parameters. Inspired by the sequential threeway decisions STWD, this paper proposes STWD with an SFNN STWDSFNN to enhance the performance of networks on structured datasets. STWDSFNN adopts multigranularity levels to dynamically learn the number of hidden layer nodes from coarse to fine, and set the sequential threshold parameters. Specifically, at the coarse granular level, STWDSFNN handles easytoclassify instances by applying strict threshold conditions, and with the increasing number of hidden layer nodes at the fine granular level, STWDSFNN focuses more on disposing of the difficulttoclassify instances by applying loose threshold conditions, thereby realizing the classification of instances. Moreover, STWDSFNN considers and reports the process cost produced from each granular level. The experimental results verify that STWDSFNN has a more compact network on structured datasets than other SFNN models, and has better generalization performance than the competitive models. All models and datasets can be downloaded from httpsgithub.comwuc567MachinelearningtreemainSTWDSFNN.

Improving 3D Imaging with PreTrained Perpendicular 2D Diffusion Models ; Diffusion models have become a popular approach for image generation and reconstruction due to their numerous advantages. However, most diffusionbased inverse problemsolving methods only deal with 2D images, and even recently published 3D methods do not fully exploit the 3D distribution prior. To address this, we propose a novel approach using two perpendicular pretrained 2D diffusion models to solve the 3D inverse problem. By modeling the 3D data distribution as a product of 2D distributions sliced in different directions, our method effectively addresses the curse of dimensionality. Our experimental results demonstrate that our method is highly effective for 3D medical image reconstruction tasks, including MRI Zaxis superresolution, compressed sensing MRI, and sparseview CT. Our method can generate highquality voxel volumes suitable for medical applications.

Generating contingency tables with fixed marginal probabilities and dependence structures described by loglinear models ; We present a method to generate contingency tables that follow loglinear models with prescribed marginal probabilities and dependence structures. We make use of loglinear Poisson regression, where the dependence structures, described using odds ratios, are implemented using an offset term. We apply this methodology to carry out simulation studies in the context of population size estimation using dual system and triple system estimators, popular in official statistics. These estimators use contingency tables that summarise the counts of elements enumerated or captured within lists that are linked. The simulation is used to investigate these estimators in the situation that the model assumptions are fulfilled, and the situation that the model assumptions are violated.

ZeroShot Contrastive Loss for TextGuided Diffusion Image Style Transfer ; Diffusion models have shown great promise in textguided image style transfer, but there is a tradeoff between style transformation and content preservation due to their stochastic nature. Existing methods require computationally expensive finetuning of diffusion models or additional neural network. To address this, here we propose a zeroshot contrastive loss for diffusion models that doesn't require additional finetuning or auxiliary networks. By leveraging patchwise contrastive loss between generated samples and original image embeddings in the pretrained diffusion model, our method can generate images with the same semantic content as the source image in a zeroshot manner. Our approach outperforms existing methods while preserving content and requiring no additional training, not only for image style transfer but also for imagetoimage translation and manipulation. Our experimental results validate the effectiveness of our proposed method.

ResDiff Combining CNN and Diffusion Model for Image SuperResolution ; Adapting the Diffusion Probabilistic Model DPM for direct image superresolution is wasteful, given that a simple Convolutional Neural Network CNN can recover the main lowfrequency content. Therefore, we present ResDiff, a novel Diffusion Probabilistic Model based on Residual structure for Single Image SuperResolution SISR. ResDiff utilizes a combination of a CNN, which restores primary lowfrequency components, and a DPM, which predicts the residual between the groundtruth image and the CNNpredicted image. In contrast to the common diffusionbased methods that directly use LR images to guide the noise towards HR space, ResDiff utilizes the CNN's initial prediction to direct the noise towards the residual space between HR space and CNNpredicted space, which not only accelerates the generation process but also acquires superior sample quality. Additionally, a frequencydomainbased loss function for CNN is introduced to facilitate its restoration, and a frequencydomain guided diffusion is designed for DPM on behalf of predicting highfrequency details. The extensive experiments on multiple benchmark datasets demonstrate that ResDiff outperforms previous diffusionbased methods in terms of shorter model convergence time, superior generation quality, and more diverse samples.

Real Options Technique as a Tool of Strategic Risk Management ; The real options approach is now considered an effective alternative to the corporate DCF model for a feasibility study. The current paper offers a practical methodology employing binomial trees and real options techniques for evaluating investment projects. A general computation procedure is suggested for the decision tree with two active stages of real options, which correspond to additional investments. The suggested technique can be used for most real options, which are practically essential regarding enterprise strategy. The special case named BinomialRandomCashFlow Real Options Model with random outcomes is developed as the next step of real options modelling. Project Value at Risk is introduced and used as a criterion of investment project feasibility under the assumption regarding random outcomes. In particular, the Gaussian probability distribution is used for modelling option outcomes uncertainty. The choice of the Gaussian distribution is caused by the desire to obtain estimates in the final analytical form. Choosing another distribution for random outcomes leads to using Monte Carlo simulation, for which a general framework is developed by demonstrating some instances. The author could avoid the computational complexity that makes these solutions feasible for business practice.

DiffusionSeg Adapting Diffusion Towards Unsupervised Object Discovery ; Learning from a large corpus of data, pretrained models have achieved impressive progress nowadays. As popular generative pretraining, diffusion models capture both lowlevel visual knowledge and highlevel semantic relations. In this paper, we propose to exploit such knowledgeable diffusion models for mainstream discriminative tasks, i.e., unsupervised object discovery saliency segmentation and object localization. However, the challenges exist as there is one structural difference between generative and discriminative models, which limits the direct use. Besides, the lack of explicitly labeled data significantly limits performance in unsupervised settings. To tackle these issues, we introduce DiffusionSeg, one novel synthesisexploitation framework containing twostage strategies. To alleviate data insufficiency, we synthesize abundant images, and propose a novel trainingfree AttentionCut to obtain masks in the first synthesis stage. In the second exploitation stage, to bridge the structural gap, we use the inversion technique, to map the given image back to diffusion features. These features can be directly used by downstream architectures. Extensive experiments and ablation studies demonstrate the superiority of adapting diffusion for unsupervised object discovery.

NoisyHate Benchmarking Content Moderation Machine Learning Models with HumanWritten Perturbations Online ; Online texts with toxic content are a threat in social media that might cause cyber harassment. Although many platforms applied measures, such as machine learningbased hatespeech detection systems, to diminish their effect, those toxic content publishers can still evade the system by modifying the spelling of toxic words. Those modified words are also known as humanwritten text perturbations. Many research works developed certain techniques to generate adversarial samples to help the machine learning models obtain the ability to recognize those perturbations. However, there is still a gap between those machinegenerated perturbations and humanwritten perturbations. In this paper, we introduce a benchmark test set containing humanwritten perturbations online for toxic speech detection models. We also recruited a group of workers to evaluate the quality of this test set and dropped lowquality samples. Meanwhile, to check if our perturbation can be normalized to its clean version, we applied spell corrector algorithms on this dataset. Finally, we test this data on stateoftheart language models, such as BERT and RoBERTa, and black box APIs, such as perspective API, to demonstrate the adversarial attack with real humanwritten perturbations is still effective.

Elastic Interaction EnergyBased Generative Model Approximation in Feature Space ; In this paper, we propose a novel approach to generative modeling using a loss function based on elastic interaction energy EIE, which is inspired by the elastic interaction between defects in crystals. The utilization of the EIEbased metric presents several advantages, including its long range property that enables consideration of global information in the distribution. Moreover, its inclusion of a selfinteraction term helps to prevent mode collapse and captures all modes of distribution. To overcome the difficulty of the relatively scattered distribution of highdimensional data, we first map the data into a latent feature space and approximate the feature distribution instead of the data distribution. We adopt the GAN framework and replace the discriminator with a feature transformation network to map the data into a latent space. We also add a stabilizing term to the loss of the feature transformation network, which effectively addresses the issue of unstable training in GANbased algorithms. Experimental results on popular datasets, such as MNIST, FashionMNIST, CIFAR10, and CelebA, demonstrate that our EIEG GAN model can mitigate mode collapse, enhance stability, and improve model performance.

Cascaded Latent Diffusion Models for HighResolution Chest Xray Synthesis ; While recent advances in largescale foundational models show promising results, their application to the medical domain has not yet been explored in detail. In this paper, we progress into the realms of largescale modeling in medical synthesis by proposing Cheff a foundational cascaded latent diffusion model, which generates highlyrealistic chest radiographs providing stateoftheart quality on a 1megapixel scale. We further propose MaCheX, which is a unified interface for public chest datasets and forms the largest open collection of chest Xrays up to date. With Cheff conditioned on radiological reports, we further guide the synthesis process over text prompts and unveil the research area of reporttochestXray generation.

Zero1to3 Zeroshot One Image to 3D Object ; We introduce Zero1to3, a framework for changing the camera viewpoint of an object given just a single RGB image. To perform novel view synthesis in this underconstrained setting, we capitalize on the geometric priors that largescale diffusion models learn about natural images. Our conditional diffusion model uses a synthetic dataset to learn controls of the relative camera viewpoint, which allow new images to be generated of the same object under a specified camera transformation. Even though it is trained on a synthetic dataset, our model retains a strong zeroshot generalization ability to outofdistribution datasets as well as inthewild images, including impressionist paintings. Our viewpointconditioned diffusion approach can further be used for the task of 3D reconstruction from a single image. Qualitative and quantitative experiments show that our method significantly outperforms stateoftheart singleview 3D reconstruction and novel view synthesis models by leveraging Internetscale pretraining.

DataDriven Exact Pole Placement for Linear Systems ; The exact pole placement problem concerns computing a feedback gain that will assign the poles of a system, controlled via static state feedback, at a set of prespecified locations. This is a classic problem in feedback control and numerous methodologies have been proposed in the literature for cases where a model of the system to control is available. In this paper, we study the problem of computing feedback gains for pole placement and, more generally, eigenstructure assignment directly from experimental data. Interestingly, we show that the closedloop poles can be placed exactly at arbitrary locations without relying on any model description but by using only finitelength trajectories generated by the openloop system. In turn, these findings imply that classical control objectives, such as feedback stabilization or meeting transient performance specifications, can be achieved without first identifying a system model. Numerical experiments demonstrate the benefits of the datadriven poleplacement approach as compared to its modelbased counterpart.

A Simple Explanation for the Phase Transition in Large Language Models with List Decoding ; Various recent experimental results show that large language models LLM exhibit emergent abilities that are not present in small models. System performance is greatly improved after passing a certain critical threshold of scale. In this letter, we provide a simple explanation for such a phase transition phenomenon. For this, we model an LLM as a sequencetosequence random function. Instead of using instant generation at each step, we use a list decoder that keeps a list of candidate sequences at each step and defers the generation of the output sequence at the end. We show that there is a critical threshold such that the expected number of erroneous candidate sequences remains bounded when an LLM is below the threshold, and it grows exponentially when an LLM is above the threshold. Such a threshold is related to the basic reproduction number in a contagious disease.

A 3catenane nonautonomous molecular motor model geometric phase, nopumping theorem, and energy transduction ; We study a model of synthetic molecular motor a 3catenane consisting of two small macrocycles mechanically interlocked with a bigger one subjected to a timedependent driving using stochastic thermodynamics. The model presents nontrivial features due to the two interacting small macrocycles, but is simple enough to be treated analytically in limiting regimes. Among the results obtained, we find a mapping into an equivalent 2catenane that reveals the implications of the nopumping theorem stating that to generate net motion of the small macrocycles, both energies and barriers need to change. In the adiabatic limit slow driving, we fully characterize the motor's dynamics and show that the net motion of the small macrocycles is expressed as a surface integral in parameter space which corrects previous erroneous results. We also analyze the performance of the motor subjected to a stepwise driving protocols in absence and in presence of an applied load. Optimization strategies for generating large currents and maximizing freeenergy transduction are proposed. This simple model provides interesting clues into the working principles of nonautonomous molecular motors and their optimization.

An AgentBased Model for Poverty and Discrimination PolicyMaking ; The deceleration of global poverty reduction in the last decades suggests that traditional redistribution policies are losing their effectiveness. Alternative ways to work towards the 1 United Nations Sustainable Development Goal poverty eradication are required. NGOs have insistingly denounced the criminalization of poverty, and the social science literature suggests that discrimination against the poor a phenomenon known as aporophobia could constitute a brake to the fight against poverty. This paper describes a proposal for an agentbased model to examine the impact that aporophobia at the institutional level has on poverty levels. This aporophobia agentbased model AABM will first be applied to a case study in the city of Barcelona. The regulatory environment is central to the model, since aporophobia has been identified in the legal framework. The AABM presented in this paper constitutes a cornerstone to obtain empirical evidence, in a noninvasive way, on the causal relationship between aporophobia and poverty levels. The simulations that will be generated based on the AABM have the potential to inform a new generation of poverty reduction policies, which act not only on the redistribution of wealth but also on the discrimination of the poor.

DiffuScene Scene Graph Denoising Diffusion Probabilistic Model for Generative Indoor Scene Synthesis ; We present DiffuScene for indoor 3D scene synthesis based on a novel scene graph denoising diffusion probabilistic model, which generates 3D instance properties stored in a fullyconnected scene graph and then retrieves the most similar object geometry for each graph node i.e. object instance which is characterized as a concatenation of different attributes, including location, size, orientation, semantic, and geometry features. Based on this scene graph, we designed a diffusion model to determine the placements and types of 3D instances. Our method can facilitate many downstream applications, including scene completion, scene arrangement, and textconditioned scene synthesis. Experiments on the 3DFRONT dataset show that our method can synthesize more physically plausible and diverse indoor scenes than stateoftheart methods. Extensive ablation studies verify the effectiveness of our design choice in scene diffusion models.

LMCanvas ObjectOriented Interaction to Personalize Large Language ModelPowered Writing Environments ; Large language models LLMs can enhance writing by automating or supporting specific tasks in writers' workflows e.g., paraphrasing, creating analogies. Leveraging this capability, a collection of interfaces have been developed that provide LLMpowered tools for specific writing tasks. However, these interfaces provide limited support for writers to create personal tools for their own unique tasks, and may not comprehensively fulfill a writer's needs requiring them to continuously switch between interfaces during writing. In this work, we envision LMCanvas, an interface that enables writers to create their own LLMpowered writing tools and arrange their personal writing environment by interacting with blocks in a canvas. In this interface, users can create text blocks to encapsulate writing and LLM prompts, model blocks for model parameter configurations, and connect these to create pipeline blocks that output generations. In this workshop paper, we discuss the design for LMCanvas and our plans to develop this concept.

Personalized Federated Learning on LongTailed Data via Adversarial Feature Augmentation ; Personalized Federated Learning PFL aims to learn personalized models for each client based on the knowledge across all clients in a privacypreserving manner. Existing PFL methods generally assume that the underlying global data across all clients are uniformly distributed without considering the longtail distribution. The joint problem of data heterogeneity and longtail distribution in the FL environment is more challenging and severely affects the performance of personalized models. In this paper, we propose a PFL method called Federated Learning with Adversarial Feature Augmentation FedAFA to address this joint problem in PFL. FedAFA optimizes the personalized model for each client by producing a balanced feature set to enhance the local minority classes. The local minority class features are generated by transferring the knowledge from the local majority class features extracted by the global model in an adversarial example learning manner. The experimental results on benchmarks under different settings of data heterogeneity and longtail distribution demonstrate that FedAFA significantly improves the personalized performance of each client compared with the stateoftheart PFL algorithm. The code is available at httpsgithub.compxqianFedAFA.

Generalized epidemic model incorporating nonMarkovian infection processes and waning immunity ; The Markovian approach, which assumes exponentially distributed interinfection times, is dominant in epidemic modeling. However, this assumption is unrealistic as an individual's infectiousness depends on its viral load and varies over time. In this paper, we present a SusceptibleInfectedRecoveredVaccinatedSusceptible epidemic model incorporating nonMarkovian infection processes. The model can be easily adapted to accurately capture the generation time distributions of emerging infectious diseases, which is essential for accurate epidemic prediction. We observe noticeable variations in the transient behavior under different infectiousness profiles and the same basic reproduction number R0. The theoretical analyses show that only R0 and the mean immunity period of the vaccinated individuals have an impact on the critical vaccination rate needed to achieve herd immunity. A vaccination level at the critical vaccination rate can ensure a very low incidence among the population in case of future epidemics, regardless of the infectiousness profiles.

Implicit Diffusion Models for Continuous SuperResolution ; Image superresolution SR has attracted increasing attention due to its wide applications. However, current SR methods generally suffer from oversmoothing and artifacts, and most work only with fixed magnifications. This paper introduces an Implicit Diffusion Model IDM for highfidelity continuous image superresolution. IDM integrates an implicit neural representation and a denoising diffusion model in a unified endtoend framework, where the implicit neural representation is adopted in the decoding process to learn continuousresolution representation. Furthermore, we design a scalecontrollable conditioning mechanism that consists of a lowresolution LR conditioning network and a scaling factor. The scaling factor regulates the resolution and accordingly modulates the proportion of the LR information and generated features in the final output, which enables the model to accommodate the continuousresolution requirement. Extensive experiments validate the effectiveness of our IDM and demonstrate its superior performance over prior arts.

Quantization of integrable and chaotic threeparticle FermiPastaUlamTsingou models ; We study the transition from integrability to chaos for the threeparticle FermiPastaUlam Tsingou FPUT model. We can show that both the quartic bFPUT model alpha 0 and the cubic one beta 0 are integrable by introducing an appropriate Fourier representation to express the nonlinear terms of the Hamiltonian. For generic values of alpha and beta, the model is nonintegrable and displays a mixed phase space with both chaotic and regular trajectories. In the classical case, chaos is diagnosed by the investigation of Poincar'e sections. In the quantum case, the level spacing statistics in the energy basis belongs to the Gaussian orthogonal ensemble in the chaotic regime, and crosses over to Poissonian behavior in the quasiintegrable lowenergy limit. In the chaotic part of the spectrum, two generic observables obey the eigenstate thermalization hypothesis.

Change Point Detection on a Separable Model for Dynamic Networks ; This paper studies the change point detection problem in time series of networks, with the Separable Temporal Exponentialfamily Random Graph Model STERGM. We consider a sequence of networks generated from a piecewise constant distribution that is altered at unknown change points in time. Detection of the change points can identify the discrepancies in the underlying data generating processes and facilitate downstream dynamic network analysis tasks. Moreover, the STERGM that focuses on network statistics is a flexible model to fit dynamic networks with both dyadic and temporal dependence. We propose a new estimator derived from the Alternating Direction Method of Multipliers ADMM and the Group Fused Lasso to simultaneously detect multiple time points, where the parameters of STERGM have changed. We also provide Bayesian information criterion for model selection to assist the detection. Our experiments show good performance of the proposed method on both simulated and real data. Lastly, we develop an R package CPDstergm to implement our method.

Ensemble Methods for MultiOrgan Segmentation in CT Series ; In the medical images field, semantic segmentation is one of the most important, yet difficult and timeconsuming tasks to be performed by physicians. Thanks to the recent advancement in the Deep Learning models regarding Computer Vision, the promise to automate this kind of task is getting more and more realistic. However, many problems are still to be solved, like the scarce availability of data and the difficulty to extend the efficiency of highly specialised models to general scenarios. Organs at risk segmentation for radiotherapy treatment planning falls in this category, as the limited data available negatively affects the possibility to develop generalpurpose models; in this work, we focus on the possibility to solve this problem by presenting three types of ensembles of singleorgan models able to produce multiorgan masks exploiting the different specialisations of their components. The results obtained are promising and prove that this is a possible solution to finding efficient multiorgan segmentation methods.

Resolving power A general approach to compare the discriminating capacity of thresholdfree evaluation metrics ; This paper introduces the concept of resolving power to describe the capacity of an evaluation metric to discriminate between models of similar quality. This capacity depends on two attributes 1. The metric's response to improvements in model quality its signal, and 2. The metric's sampling variability its noise. The paper defines resolving power as a metric's sampling uncertainty scaled by its signal. Resolving power's primary application is to compare the discriminating capacity of thresholdfree evaluation metrics, such as the area under the receiver operating characteristic curve AUROC and the area under the precisionrecall curve AUPRC. A simulation study compares the AUROC and the AUPRC in a variety of contexts. The analysis suggests that the AUROC generally has greater resolving power, but that the AUPRC is superior in some conditions, such as those where highquality models are applied to low prevalence outcomes. The paper concludes by proposing an empirical method to estimate resolving power that can be applied to any dataset and any initial classification model.

Towards SelfExplainability of Deep Neural Networks with Heatmap Captioning and LargeLanguage Models ; Heatmaps are widely used to interpret deep neural networks, particularly for computer vision tasks, and the heatmapbased explainable AI XAI techniques are a wellresearched topic. However, most studies concentrate on enhancing the quality of the generated heatmap or discovering alternate heatmap generation techniques, and little effort has been devoted to making heatmapbased XAI automatic, interactive, scalable, and accessible. To address this gap, we propose a framework that includes two modules 1 context modelling and 2 reasoning. We proposed a templatebased image captioning approach for context modelling to create textbased contextual information from the heatmap and input data. The reasoning module leverages a large language model to provide explanations in combination with specialised knowledge. Our qualitative experiments demonstrate the effectiveness of our framework and heatmap captioning approach. The code for the proposed templatebased heatmap captioning approach will be publicly available.

Fewshot Semantic Image Synthesis with Class Affinity Transfer ; Semantic image synthesis aims to generate photo realistic images given a semantic segmentation map. Despite much recent progress, training them still requires large datasets of images annotated with perpixel label maps that are extremely tedious to obtain. To alleviate the high annotation cost, we propose a transfer method that leverages a model trained on a large source dataset to improve the learning ability on small target datasets via estimated pairwise relations between source and target classes. The class affinity matrix is introduced as a first layer to the source model to make it compatible with the target label maps, and the source model is then further finetuned for the target domain. To estimate the class affinities we consider different approaches to leverage prior knowledge semantic segmentation on the source domain, textual label embeddings, and selfsupervised vision features. We apply our approach to GANbased and diffusionbased architectures for semantic synthesis. Our experiments show that the different ways to estimate class affinity can be effectively combined, and that our approach significantly improves over existing stateoftheart transfer approaches for generative image models.

CrossCarry An R package for the analysis of data from a crossover design with GEE ; Experimental crossover designs are widely used in medicine, agriculture, and other areas of the biological sciences. Due to the characteristics of the crossover design, each experimental unit has longitudinal observations and the presence of drag effects on the response variable. There is no package in R that clearly models data from crossover designs. The CrossCarry package presented in this paper allows testing any crossover design as long as the observed response variable belongs to the exponential family, regardless of whether or not there is a washout period. It also allows modeling repeated measurements within each period and extends the correlation structures used in the generalized estimating equations. The family of correlation structures is built that takes into account the particularities of the design, that is, the correlation between and within the periods. It also includes a parametric component for modeling treatment effects and a nonparametric component for modeling time effects and carryover effects. The nonparametric component is estimated from splines inserted into the generalized estimation equations.

A review of ensemble learning and data augmentation models for class imbalanced problems combination, implementation and evaluation ; Class imbalance CI in classification problems arises when the number of observations belonging to one class is lower than the other. Ensemble learning combines multiple models to obtain a robust model and has been prominently used with data augmentation methods to address class imbalance problems. In the last decade, a number of strategies have been added to enhance ensemble learning and data augmentation methods, along with new methods such as generative adversarial networks GANs. A combination of these has been applied in many studies, but the true rank of different combinations would require a computational review. In this paper, we present a computational review to evaluate data augmentation and ensemble learning methods used to address prominent benchmark CI problems. We present a general framework that evaluates 10 data augmentation and 10 ensemble learning methods for CI problems. Our objective is to identify the most effective combination for improving classification performance on imbalanced datasets. The results indicate that combinations of data augmentation methods with ensemble learning can significantly improve classification performance on imbalanced datasets. Our study is vital for the development of novel models for handling imbalanced datasets.

Causal Discovery with Score Matching on Additive Models with Arbitrary Noise ; Causal discovery methods are intrinsically constrained by the set of assumptions needed to ensure structure identifiability. Moreover additional restrictions are often imposed in order to simplify the inference task this is the case for the Gaussian noise assumption on additive nonlinear models, which is common to many causal discovery approaches. In this paper we show the shortcomings of inference under this hypothesis, analyzing the risk of edge inversion under violation of Gaussianity of the noise terms. Then, we propose a novel method for inferring the topological ordering of the variables in the causal graph, from data generated according to an additive nonlinear model with a generic noise distribution. This leads to NoGAM Not only Gaussian Additive noise Models, a causal discovery algorithm with a minimal set of assumptions and state of the art performance, experimentally benchmarked on synthetic data.

Towards Unified Scene Text Spotting based on Sequence Generation ; Sequence generation models have recently made significant progress in unifying various vision tasks. Although some autoregressive models have demonstrated promising results in endtoend text spotting, they use specific detection formats while ignoring various text shapes and are limited in the maximum number of text instances that can be detected. To overcome these limitations, we propose a UNIfied scene Text Spotter, called UNITS. Our model unifies various detection formats, including quadrilaterals and polygons, allowing it to detect text in arbitrary shapes. Additionally, we apply startingpoint prompting to enable the model to extract texts from an arbitrary starting point, thereby extracting more texts beyond the number of instances it was trained on. Experimental results demonstrate that our method achieves competitive performance compared to stateoftheart methods. Further analysis shows that UNITS can extract a larger number of texts than it was trained on. We provide the code for our method at httpsgithub.comclovaaiunits.

Weakly supervised segmentation with point annotations for histopathology images via contrastbased variational model ; Image segmentation is a fundamental task in the field of imaging and vision. Supervised deep learning for segmentation has achieved unparalleled success when sufficient training data with annotated labels are available. However, annotation is known to be expensive to obtain, especially for histopathology images where the target regions are usually with high morphology variations and irregular shapes. Thus, weakly supervised learning with sparse annotations of points is promising to reduce the annotation workload. In this work, we propose a contrastbased variational model to generate segmentation results, which serve as reliable complementary supervision to train a deep segmentation model for histopathology images. The proposed method considers the common characteristics of target regions in histopathology images and can be trained in an endtoend manner. It can generate more regionally consistent and smoother boundary segmentation, and is more robust to unlabeled novel' regions. Experiments on two different histology datasets demonstrate its effectiveness and efficiency in comparison to previous models.

Synthetic turbulence generator for the wallmodeled LES lattice Boltzmann method ; The synthetic turbulence generator STG lies at the interface of the Reynolds averaged NavierStokes RANS simulation and large eddy simulation LES. This paper presents a STG for the multiplerelaxationtimeMRT lattice Boltzmann methodLBM framework at high friction Reynolds numbers, with consideration of near wall modeling. The Reichardt wall law, in combination with a forcebased method, is used to model the near wall field. The STG wallmodeledSTGWM LES results are compared with turbulent channel flow simulations at Retau1000,2000,5200 at different resolutions. The results demonstrate good agreement with DNS, with the adaptation length of 6 to 8 boundary layer thickness. This method has a wide range of potentials for hybrid RANSLESLBM related applications at high friction Reynolds numbers.

On the IRUV flavour connection in nonuniversal axion models ; Nonuniversal axion models, with the PecceiQuinn PQ symmetry acting on Standard Model SM fermions in a generationdependent way, are typically accompanied by two different sources of flavour violation, dubbed here as infrared IR and ultraviolet UV. The former is due to the flavour violating axion couplings to SM fermions, while the latter arises from the heavy degrees of freedom that UV complete the axion effective field theory. We point out that these two sources of flavour violation are directly related and exemplify this connection in a general class of nonuniversal axion model, based on a renormalizable DFSZlike setup with two Higgs doublets PQ2HDM. We next discuss the interplay of axion flavour phenomenology with the signatures stemming from the heavy radial modes of the PQ2HDM, including meson oscillation observables and charged lepton flavour violating decays. We emphasize the strong complementarity between flavour observables, LHC direct searches and standard axion physics.

Reflected Diffusion Models ; Scorebased diffusion models learn to reverse a stochastic differential equation that maps data to noise. However, for complex tasks, numerical error can compound and result in highly unnatural samples. Previous work mitigates this drift with thresholding, which projects to the natural data domain such as pixel space for images after each diffusion step, but this leads to a mismatch between the training and generative processes. To incorporate data constraints in a principled manner, we present Reflected Diffusion Models, which instead reverse a reflected stochastic differential equation evolving on the support of the data. Our approach learns the perturbed score function through a generalized score matching loss and extends key components of standard diffusion models including diffusion guidance, likelihoodbased training, and ODE sampling. We also bridge the theoretical gap with thresholding such schemes are just discretizations of reflected SDEs. On standard image benchmarks, our method is competitive with or surpasses the state of the art without architectural modifications and, for classifierfree guidance, our approach enables fast exact sampling with ODEs and produces more faithful samples under high guidance weight.

Neural Multinetwork Diffusion towards Social Recommendation ; Graph Neural Networks GNNs have been widely applied on a variety of realworld applications, such as social recommendation. However, existing GNNbased models on social recommendation suffer from serious problems of generalization and oversmoothness, because of the underexplored negative sampling method and the direct implanting of the offtheshelf GNN models. In this paper, we propose a succinct multinetwork GNNbased neural model NeMo for social recommendation. Compared with the existing methods, the proposed model explores a generative negative sampling strategy, and leverages both the positive and negative useritem interactions for users' interest propagation. The experiments show that NeMo outperforms the stateoftheart baselines on various realworld benchmark datasets e.g., by up to 38.8 in terms of NDCG15.

The No Free Lunch Theorem, Kolmogorov Complexity, and the Role of Inductive Biases in Machine Learning ; No free lunch theorems for supervised learning state that no learner can solve all problems or that all learners achieve exactly the same accuracy on average over a uniform distribution on learning problems. Accordingly, these theorems are often referenced in support of the notion that individual problems require specially tailored inductive biases. While virtually all uniformly sampled datasets have high complexity, realworld problems disproportionately generate lowcomplexity data, and we argue that neural network models share this same preference, formalized using Kolmogorov complexity. Notably, we show that architectures designed for a particular domain, such as computer vision, can compress datasets on a variety of seemingly unrelated domains. Our experiments show that pretrained and even randomly initialized language models prefer to generate lowcomplexity sequences. Whereas no free lunch theorems seemingly indicate that individual problems require specialized learners, we explain how tasks that often require human intervention such as picking an appropriately sized model when labeled data is scarce or plentiful can be automated into a single learning algorithm. These observations justify the trend in deep learning of unifying seemingly disparate problems with an increasingly small set of machine learning models.

ImageReward Learning and Evaluating Human Preferences for TexttoImage Generation ; We present a comprehensive solution to learn and improve texttoimage models from human preference feedback. To begin with, we build ImageReward the first generalpurpose texttoimage human preference reward model to effectively encode human preferences. Its training is based on our systematic annotation pipeline including rating and ranking, which collects 137k expert comparisons to date. In human evaluation, ImageReward outperforms existing scoring models and metrics, making it a promising automatic metric for evaluating texttoimage synthesis. On top of it, we propose Reward Feedback Learning ReFL, a direct tuning algorithm to optimize diffusion models against a scorer. Both automatic and human evaluation support ReFL's advantages over compared methods. All code and datasets are provided at urlhttpsgithub.comTHUDMImageReward.

Study of Decoupled Gravastars in Energymomentum Squared Gravity ; In this paper, we generate an exact anisotropic gravastar model using gravitational decoupling technique through minimal geometric deformation in the framework of fRe, T2 gravity. This novel model explains an ultracompact stellar configuration whose internal region is smoothly matched to the exterior region. The developed stellar model satisfies some of the essential characteristics of a physically acceptable model such as a positive monotonically decreasing profile of energy density from the center to the boundary and monotonically decreasing behavior of the pressure. The anisotropic factor and Schwarzschild spacetime follows physically acceptable behavior. We find that all the energy bounds are satisfied except strong energy condition inside the ultracompact stellar structure for the coupling constant of this theory, which is compatible with the regularity condition.

Kinematic Earthquake Sequences on Geometrically Complex Faults ; Computational earthquake sequence models provide generative estimates of the time, location, and size of synthetic seismic events that can be compared with observed earthquake histories and assessed as rupture forecasts. Here we describe a threedimensional probabilistic earthquake sequence model that produces slip event time series constrained across geometrically complex nonplanar fault systems. This model is kinematic in nature, integrating the time evolution of geometric moment accumulation and release with empirical earthquake scaling laws. The temporal probability of event occurrence is determined from the time history of geometric moment integrated with shortterm Omoristyle rate decay following each earthquake achieving longterm timeaveraged moment balance. Similarly, the net geometric moment monotonically controls the probability of event localization, and seismic events release geometric moment with spatially heterogenous slip on threedimensional nonplanar fault surfaces. We use this model to generate a synthetic earthquake sequence on the Nankai subduction zone over a 1,250yearlong interval, including 700 MW5.58.5 coseismic events, with decadaltocentennial scale quiescent intervals quasiperiodic great earthquake clusters followed by aftershock sequences.

Learning SemanticAware Knowledge Guidance for LowLight Image Enhancement ; Lowlight image enhancement LLIE investigates how to improve illumination and produce normallight images. The majority of existing methods improve lowlight images via a global and uniform manner, without taking into account the semantic information of different regions. Without semantic priors, a network may easily deviate from a region's original color. To address this issue, we propose a novel semanticaware knowledgeguided framework SKF that can assist a lowlight enhancement model in learning rich and diverse priors encapsulated in a semantic segmentation model. We concentrate on incorporating semantic knowledge from three key aspects a semanticaware embedding module that wisely integrates semantic priors in feature representation space, a semanticguided color histogram loss that preserves color consistency of various instances, and a semanticguided adversarial loss that produces more natural textures by semantic priors. Our SKF is appealing in acting as a general framework in LLIE task. Extensive experiments show that models equipped with the SKF significantly outperform the baselines on multiple datasets and our SKF generalizes to different models and scenes well. The code is available at SemanticAwareLowLightImageEnhancement.