Datasets:

sswt

/

arxiv_sample_50K

like 0

Modeling Quasar UVOptical Variability with the Coronaheated Accretiondisk Reprocessing CHAR Model ; The restframe UVoptical variability of the quasars in the Sloan Digital Sky Survey SDSS Stripe 82 is used to test the CoronaHeated Accretiondisk Reprocessing CHAR model of Sun et al. 2020. We adopt our CHAR model and the observed blackhole masses MmathrmBH and luminosities L to generate mock light curves that share the same measurement noise and sampling as the real observations. Without any finetuning, our CHAR model can satisfactorily reproduce the observed ensemble structure functions for different MmathrmBH, L, and restframe wavelengths. Our analyses reveal that a luminositydependent bolometric correction is disfavored over the constant bolometric correction for UVoptical luminosities. Our work demonstrates the possibility of extracting quasar properties e.g., the bolometric correction or the dimensionless viscosity parameter by comparing the physical CHAR model with quasar light curves.

Importance subsampling for power system planning under multiyear demand and weather uncertainty ; This paper introduces a generalised version of importance subsampling for time series reductionaggregation in optimisationbased power system planning models. Recent studies indicate that reliably determining optimal electricity investment strategy under climate variability requires the consideration of multiple years of demand and weather data. However, solving planning models over long simulation lengths is typically computationally unfeasible, and established time series reduction approaches induce significant errors. The importance subsampling method reliably estimates longterm planning model outputs at greatly reduced computational cost, allowing the consideration of multidecadal samples. The key innovation is a systematic identification and preservation of relevant extreme events in modeling subsamples. Simulation studies on generation and transmission expansion planning models illustrate the method's enhanced performance over established representative days clustering approaches. The models, data and sample code are made available as opensource software.

Smart Weather Forecasting Using Machine LearningA Case Study in Tennessee ; Traditionally, weather predictions are performed with the help of large complex models of physics, which utilize different atmospheric conditions over a long period of time. These conditions are often unstable because of perturbations of the weather system, causing the models to provide inaccurate forecasts. The models are generally run on hundreds of nodes in a large High Performance Computing HPC environment which consumes a large amount of energy. In this paper, we present a weather prediction technique that utilizes historical data from multiple weather stations to train simple machine learning models, which can provide usable forecasts about certain weather conditions for the near future within a very short period of time. The models can be run on much less resource intensive environments. The evaluation results show that the accuracy of the models is good enough to be used alongside the current stateoftheart techniques. Furthermore, we show that it is beneficial to leverage the weather station data from multiple neighboring areas over the data of only the area for which weather forecasting is being performed.

Unbiased estimator for the variance of the leaveoneout crossvalidation estimator for a Bayesian normal model with fixed variance ; When evaluating and comparing models using leaveoneout crossvalidation LOOCV, the uncertainty of the estimate is typically assessed using the variance of the sampling distribution. Considering the uncertainty is important, as the variability of the estimate can be high in some cases. An important result by Bengio and Grandvalet 2004 states that no general unbiased variance estimator can be constructed, that would apply for any utility or loss measure and any model. We show that it is possible to construct an unbiased estimator considering a specific predictive performance measure and model. We demonstrate an unbiased sampling distribution variance estimator for the Bayesian normal model with fixed model variance using the expected log pointwise predictive density elpd utility score. This example demonstrates that it is possible to obtain improved, problemspecific, unbiased estimators for assessing the uncertainty in LOOCV estimation.

Improved Weighted Random Forest for Classification Problems ; Several studies have shown that combining machine learning models in an appropriate way will introduce improvements in the individual predictions made by the base models. The key to make wellperforming ensemble model is in the diversity of the base models. Of the most common solutions for introducing diversity into the decision trees are bagging and random forest. Bagging enhances the diversity by sampling with replacement and generating many training data sets, while random forest adds selecting a random number of features as well. This has made the random forest a winning candidate for many machine learning applications. However, assuming equal weights for all base decision trees does not seem reasonable as the randomization of sampling and input feature selection may lead to different levels of decisionmaking abilities across base decision trees. Therefore, we propose several algorithms that intend to modify the weighting strategy of regular random forest and consequently make better predictions. The designed weighting frameworks include optimal weighted random forest based on accuracy, optimal weighted random forest based on the area under the curve AUC, performancebased weighted random forest, and several stackingbased weighted random forest models. The numerical results show that the proposed models are able to introduce significant improvements compared to regular random forest.

A Comparison of Pretrained VisionandLanguage Models for Multimodal Representation Learning across Medical Images and Reports ; Joint imagetext embedding extracted from medical images and associated contextual reports is the bedrock for most biomedical visionandlanguage VL tasks, including medical visual question answering, clinical imagetext retrieval, clinical report autogeneration. In this study, we adopt four pretrained VL models LXMERT, VisualBERT, UNIER and PixelBERT to learn multimodal representation from MIMICCXR radiographs and associated reports. The extrinsic evaluation on OpenI dataset shows that in comparison to the pioneering CNNRNN model, the joint embedding learned by pretrained VL models demonstrate performance improvement in the thoracic findings classification task. We conduct an ablation study to analyze the contribution of certain model components and validate the advantage of joint embedding over textonly embedding. We also visualize attention maps to illustrate the attention mechanism of VL models.

DataDriven Power Electronic Converter Modeling for Low Inertia Power System Dynamic Studies ; A significant amount of converterbased generation is being integrated into the bulk electric power grid to fulfill the future electric demand through renewable energy sources, such as wind and photovoltaic. The dynamics of converter systems in the overall stability of the power system can no longer be neglected as in the past. Numerous efforts have been made in the literature to derive detailed dynamic models, but using detailed models becomes complicated and computationally prohibitive in large system level studies. In this paper, we use a datadriven, blackbox approach to model the dynamics of a power electronic converter. System identification tools are used to identify the dynamic models, while a power amplifier controlled by a realtime digital simulator is used to perturb and control the converter. A set of linear dynamic models for the converter are derived, which can be employed for system level studies of converterdominated electric grids.

Graphical Gaussian Process Models for Highly Multivariate Spatial Data ; For multivariate spatial Gaussian process GP models, customary specifications of crosscovariance functions do not exploit relational intervariable graphs to ensure processlevel conditional independence among the variables. This is undesirable, especially for highly multivariate settings, where popular crosscovariance functions such as the multivariate Mat'ern suffer from a curse of dimensionality as the number of parameters and floating point operations scale up in quadratic and cubic order, respectively, in the number of variables. We propose a class of multivariate Graphical Gaussian Processes using a general construction called stitching that crafts crosscovariance functions from graphs and ensures processlevel conditional independence among variables. For the Mat'ern family of functions, stitching yields a multivariate GP whose univariate components are Mat'ern GPs, and conforms to processlevel conditional independence as specified by the graphical model. For highly multivariate settings and decomposable graphical models, stitching offers massive computational gains and parameter dimension reduction. We demonstrate the utility of the graphical Mat'ern GP to jointly model highly multivariate spatial data using simulation examples and an application to airpollution modelling.

Distinct Critical Behaviors from the Same State in Quantum Spin and Population Dynamics Perspectives ; There is a deep connection between the ground states of transversefield spin systems and the latetime distributions of evolving viral populations within simple models, both are obtained from the principal eigenvector of the same matrix. However, that vector is the wavefunction amplitude in the quantum spin model, whereas it is the probability itself in the population model. We show that this seemingly minor difference has significant consequences phase transitions which are discontinuous in the spin system become continuous when viewed through the population perspective, and transitions which are continuous become governed by new critical exponents. We introduce a more general class of models which encompasses both cases, and that can be solved exactly in a meanfield limit. Numerical results are also presented for a number of onedimensional chains with powerlaw interactions. We see that wellworn spin models of quantum statistical mechanics can contain unexpected new physics and insights when treated as populationdynamical models and beyond, motivating further studies.

An analysis of deep neural networks for predicting trends in time series data ; Recently, a hybrid Deep Neural Network DNN algorithm, TreNet was proposed for predicting trends in time series data. While TreNet was shown to have superior performance for trend prediction to other DNN and traditional ML approaches, the validation method used did not take into account the sequential nature of time series data sets and did not deal with model update. In this research we replicated the TreNet experiments on the same data sets using a walkforward validation method and tested our optimal model over multiple independent runs to evaluate model stability. We compared the performance of the hybrid TreNet algorithm, on four data sets to vanilla DNN algorithms that take in point data, and also to traditional ML algorithms. We found that in general TreNet still performs better than the vanilla DNN models, but not on all data sets as reported in the original TreNet study. This study highlights the importance of using an appropriate validation method and evaluating model stability for evaluating and developing machine learning models for trend prediction in time series data.

Group Fairness by Probabilistic Modeling with Latent Fair Decisions ; Machine learning systems are increasingly being used to make impactful decisions such as loan applications and criminal justice risk assessments, and as such, ensuring fairness of these systems is critical. This is often challenging as the labels in the data are biased. This paper studies learning fair probability distributions from biased data by explicitly modeling a latent variable that represents a hidden, unbiased label. In particular, we aim to achieve demographic parity by enforcing certain independencies in the learned model. We also show that group fairness guarantees are meaningful only if the distribution used to provide those guarantees indeed captures the realworld data. In order to closely model the data distribution, we employ probabilistic circuits, an expressive and tractable probabilistic model, and propose an algorithm to learn them from incomplete data. We evaluate our approach on a synthetic dataset in which observed labels indeed come from fair labels but with added bias, and demonstrate that the fair labels are successfully retrieved. Moreover, we show on realworld datasets that our approach not only is a better model than existing methods of how the data was generated but also achieves competitive accuracy.

Tradeoffs in Sentence Selection Techniques for OpenDomain Question Answering ; Current methods in opendomain question answering QA usually employ a pipeline of first retrieving relevant documents, then applying strong reading comprehension RC models to that retrieved text. However, modern RC models are complex and expensive to run, so techniques to prune the space of retrieved text are critical to allow this approach to scale. In this paper, we focus on approaches which apply an intermediate sentence selection step to address this issue, and investigate the best practices for this approach. We describe two groups of models for sentence selection QAbased approaches, which run a fullfledged QA system to identify answer candidates, and retrievalbased models, which find parts of each passage specifically related to each question. We examine tradeoffs between processing speed and task performance in these two approaches, and demonstrate an ensemble module that represents a hybrid of the two. From experiments on OpenSQuAD and TriviaQA, we show that very lightweight QA models can do well at this task, but retrievalbased models are faster still. An ensemble module we describe balances between the two and generalizes well crossdomain.

Predicting Geographic Information with Neural Cellular Automata ; This paper presents a novel framework using neural cellular automata NCA to regenerate and predict geographic information. The model extends the idea of using NCA to generateregenerate a specific image by training the model with various geographic data, and thus, taking the traffic condition map as an example, the model is able to predict traffic conditions by giving certain induction information. Our research verified the analogy between NCA and gene in biology, while the innovation of the model significantly widens the boundary of possible applications based on NCAs. From our experimental results, the model shows great potentials in its usability and versatility which are not available in previous studies. The code for model implementation is available at httpsredacted.

Modeling Score Distributions and Continuous Covariates A Bayesian Approach ; Computer Vision practitioners must thoroughly understand their model's performance, but conditional evaluation is complex and errorprone. In biometric verification, model performance over continuous covariatesrealnumber attributes of images that affect performanceis particularly challenging to study. We develop a generative model of the match and nonmatch score distributions over continuous covariates and perform inference with modern Bayesian methods. We use mixture models to capture arbitrary distributions and local basis functions to capture nonlinear, multivariate trends. Three experiments demonstrate the accuracy and effectiveness of our approach. First, we study the relationship between age and face verification performance and find previous methods may overstate performance and confidence. Second, we study preprocessing for CNNs and find a highly nonlinear, multivariate surface of model performance. Our method is accurate and data efficient when evaluated against previous synthetic methods. Third, we demonstrate the novel application of our method to pedestrian tracking and calculate variable thresholds and expected performance while controlling for multiple covariates.

Keeping Up Appearances Computational Modeling of Face Acts in Persuasion Oriented Discussions ; The notion of face refers to the public selfimage of an individual that emerges both from the individual's own actions as well as from the interaction with others. Modeling face and understanding its state changes throughout a conversation is critical to the study of maintenance of basic human needs in and through interaction. Grounded in the politeness theory of Brown and Levinson 1978, we propose a generalized framework for modeling face acts in persuasion conversations, resulting in a reliable coding manual, an annotated corpus, and computational models. The framework reveals insights about differences in face act utilization between asymmetric roles in persuasion conversations. Using computational models, we are able to successfully identify face acts as well as predict a key conversational outcome e.g. donation success. Finally, we model a latent representation of the conversational state to analyze the impact of predicted face acts on the probability of a positive conversational outcome and observe several correlations that corroborate previous findings.

Toward a Thermodynamics of Meaning ; As language models such as GPT3 become increasingly successful at generating realistic text, questions about what purely textbased modeling can learn about the world have become more urgent. Is text purely syntactic, as skeptics argue Or does it in fact contain some semantic information that a sufficiently sophisticated language model could use to learn about the world without any additional inputs This paper describes a new model that suggests some qualified answers to those questions. By theorizing the relationship between text and the world it describes as an equilibrium relationship between a thermodynamic system and a much larger reservoir, this paper argues that even very simple language models do learn structural facts about the world, while also proposing relatively precise limits on the nature and extent of those facts. This perspective promises not only to answer questions about what language models actually learn, but also to explain the consistent and surprising success of cooccurrence prediction as a meaningmaking strategy in AI.

Neural Twins Talk ; Inspired by how the human brain employs more neural pathways when increasing the focus on a subject, we introduce a novel twin cascaded attention model that outperforms a stateoftheart image captioning model that was originally implemented using one channel of attention for the visual grounding task. Visual grounding ensures the existence of words in the caption sentence that are grounded into a particular region in the input image. After a deep learning model is trained on visual grounding task, the model employs the learned patterns regarding the visual grounding and the order of objects in the caption sentences, when generating captions. We report the results of our experiments in three image captioning tasks on the COCO dataset. The results are reported using standard image captioning metrics to show the improvements achieved by our model over the previous image captioning model. The results gathered from our experiments suggest that employing more parallel attention pathways in a deep neural network leads to higher performance. Our implementation of NTT is publicly available at httpsgithub.comzanyarzNeuralTwinsTalk.

Reinforcement Learningbased Nary CrossSentence Relation Extraction ; The models of nary cross sentence relation extraction based on distant supervision assume that consecutive sentences mentioning n entities describe the relation of these n entities. However, on one hand, this assumption introduces noisy labeled data and harms the models' performance. On the other hand, some nonconsecutive sentences also describe one relation and these sentences cannot be labeled under this assumption. In this paper, we relax this strong assumption by a weaker distant supervision assumption to address the second issue and propose a novel sentence distribution estimator model to address the first problem. This estimator selects correctly labeled sentences to alleviate the effect of noisy data is a twolevel agent reinforcement learning model. In addition, a novel universal relation extractor with a hybrid approach of attention mechanism and PCNN is proposed such that it can be deployed in any tasks, including consecutive and nonconsecutive sentences. Experiments demonstrate that the proposed model can reduce the impact of noisy data and achieve better performance on general nary cross sentence relation extraction task compared to baseline models.

Accelerating MultiModel Inference by Merging DNNs of Different Weights ; Standardized DNN models that have been proved to perform well on machine learning tasks are widely used and often adopted asis to solve downstream tasks, forming the transfer learning paradigm. However, when serving multiple instances of such DNN models from a cluster of GPU servers, existing techniques to improve GPU utilization such as batching are inapplicable because models often do not share weights due to finetuning. We propose NetFuse, a technique of merging multiple DNN models that share the same architecture but have different weights and different inputs. NetFuse is made possible by replacing operations with more general counterparts that allow a set of weights to be associated with only a certain set of inputs. Experiments on ResNet50, ResNeXt50, BERT, and XLNet show that NetFuse can speed up DNN inference time up to 3.6x on a NVIDIA V100 GPU, and up to 3.0x on a TITAN Xp GPU when merging 32 model instances, while only using up a small additional amount of GPU memory.

Transparency, Auditability and eXplainability of Machine Learning Models in Credit Scoring ; A major requirement for credit scoring models is to provide a maximally accurate risk prediction. Additionally, regulators demand these models to be transparent and auditable. Thus, in credit scoring, very simple predictive models such as logistic regression or decision trees are still widely used and the superior predictive power of modern machine learning algorithms cannot be fully leveraged. Significant potential is therefore missed, leading to higher reserves or more credit defaults. This paper works out different dimensions that have to be considered for making credit scoring models understandable and presents a framework for making black box'' machine learning models transparent, auditable and explainable. Following this framework, we present an overview of techniques, demonstrate how they can be applied in credit scoring and how results compare to the interpretability of score cards. A real world case study shows that a comparable degree of interpretability can be achieved while machine learning techniques keep their ability to improve predictive power.

Hybrid European MVLV Network Models for Smart Distribution Network Modelling ; A pair of Europeanstyle, integrated MVLV circuits are presented, created by combining generic MV and real LV networks. The two models have 86,000 and 113,000 nodes, and are made readily available for download in the OpenDSS file format. Primary substation tap change controls and MVLV feeders are represented as threephase unbalanced distribution network models, capturing the coupling of voltages at the MV level. The assumptions made in constructing the models are outlined, including a preconditioning step that reduces the number of nodes by more than five times without affecting the solution. Two flexibilitybased case studies are presented, with TSODSO and peerpeerbased smart controls considered. The demonstration of the heterogeneous nature of these systems is corroborated by the analysis of measured LV voltage data. The models are intended to aid the development of algorithms for maximising the benefits of smart devices within the context of whole energy systems.

Do Question Answering Modeling Improvements Hold Across Benchmarks ; Do question answering QA modeling improvements e.g., choice of architecture and training procedure hold consistently across the diverse landscape of QA benchmarks To study this question, we introduce the notion of concurrence two benchmarks have high concurrence on a set of modeling approaches if they rank the modeling approaches similarly. We measure the concurrence between 32 QA benchmarks on a set of 20 diverse modeling approaches and find that humanconstructed benchmarks have high concurrence amongst themselves, even if their passage and question distributions are very different. Surprisingly, even downsampled humanconstructed benchmarks i.e., collecting less data and programmaticallygenerated benchmarks e.g., clozeformatted examples have high concurrence with humanconstructed benchmarks. These results indicate that, despite years of intense community focus on a small number of benchmarks, the modeling improvements studied hold broadly.

Canonical Form of Lyapunov Second Method in Mathematical Modelling and Control Design ; The objective of the paper is to put canonical Lyapunov functionCLF, canonizing diffeomorphism CD and canonical form of dynamical systems CFDS, which have led to the generalization of the Lyapunov second method, in perspective of their high efficiency for Mathematical Modelling and Control Design. We show how the symbiosis of the ideas of Henri Poincare and Nikolay Chetaev leads us to CD, CFDS and CLF. Our approach successfully translates into mathematical modelling and control design for special twoangles synchronized longitudinal maneuvering of a thrustvectored aircraft. The essentially nonlinear fivedimensional mathematical model of the longitudinal flight dynamics of a thrustvectored aircraft in a wingbody coordinate system with two controls, namely the angular deflections of a movable horizontal stabilizer and a turbojet engine nozzle, is investigated. The widesense robust and stable in the large tracking control law is designed. Its core is the hierarchical cascade of two controlling attractormediators and two controlling terminal attractors embedded in the extended phase space of the mathematical model of the aircraft longitudinal motion. The detailed demonstration of the elaborated technique of designing widesense robust tracking control for the nonlinear multidimensional mathematical model constitutes the quintessence of the paper.

Neural Recursive Belief States in MultiAgent Reinforcement Learning ; In multiagent reinforcement learning, the problem of learning to act is particularly difficult because the policies of coplayers may be heavily conditioned on information only observed by them. On the other hand, humans readily form beliefs about the knowledge possessed by their peers and leverage beliefs to inform decisionmaking. Such abilities underlie individual success in a wide range of Markov games, from bluffing in Poker to conditional cooperation in the Prisoner's Dilemma, to conventionbuilding in Bridge. Classical methods are usually not applicable to complex domains due to the intractable nature of hierarchical beliefs i.e. beliefs of other agents' beliefs. We propose a scalable method to approximate these belief structures using recursive deep generative models, and to use the belief models to obtain representations useful to acting in complex tasks. Our agents trained with belief models outperform modelfree baselines with equivalent representational capacity using common training paradigms. We also show that higherorder belief models outperform agents with lowerorder models.

229mTh isomer from a nuclear model perspective ; The physical conditions for the emergence of the extremely lowlying nuclear isomer 229mTh at approximately 8 eV are investigated in the framework of our recently proposed nuclear structure model. Our theoretical approach explains the 229mThisomer phenomenon as the result of a very fine interplay between collective quadrupoleoctupole and singleparticle dynamics in the nucleus. We find that the isomeric state can only appear in a rather limited model space of quadrupoleoctupole deformations in the singleparticle potential, with the octupole deformation being of a crucial importance for its formation. Within this deformation space the modeldescribed quantities exhibit a rather smooth behaviour close to the line of isomerground state quasidegeneracy determined by the crossing of the corresponding singleparticle orbitals. Our comprehensive analysis confirms the previous model predictions for reduced transition probabilities and the isomer magnetic moment, while showing a possibility for limited variation in the groundstate magnetic moment theoretical value. These findings prove the reliability of the model and suggest that the same dynamical mechanism could manifest in other actinide nuclei giving a general prescription for the search and exploration of similar isomer phenomena.

Invertible DenseNets with Concatenated LipSwish ; We introduce Invertible Dense Networks iDenseNets, a more parameter efficient extension of Residual Flows. The method relies on an analysis of the Lipschitz continuity of the concatenation in DenseNets, where we enforce invertibility of the network by satisfying the Lipschitz constant. Furthermore, we propose a learnable weighted concatenation, which not only improves the model performance but also indicates the importance of the concatenated weighted representation. Additionally, we introduce the Concatenated LipSwish as activation function, for which we show how to enforce the Lipschitz condition and which boosts performance. The new architecture, iDenseNet, outperforms Residual Flow and other flowbased models on density estimation evaluated in bits per dimension, where we utilize an equal parameter budget. Moreover, we show that the proposed model outperforms Residual Flows when trained as a hybrid model where the model is both a generative and a discriminative model.

Invariants for level1 phylogenetic networks under the CavendarFarrisNeyman Model ; Phylogenetic networks can model more complicated evolutionary phenomena that trees fail to capture such as horizontal gene transfer and hybridization. The same Markov models that are used to model evolution on trees can also be extended to networks and similar questions, such as the identifiability of the network parameter or the invariants of the model, can be asked. In this paper we focus on finding the invariants of the CavendarFarrisNeyman CFN model on level1 phylogenetic networks. We do this by reducing the problem to finding invariants of sunlet networks, which are level1 networks consisting of a single cycle with leaves at each vertex. We then determine all quadratic invariants in the sunlet network ideal which we conjecture generate the full ideal.

Rotating shallow water flow under location uncertainty with a structurepreserving discretization ; We introduce a physically relevant stochastic representation of the rotating shallow water equations. The derivation relies mainly on a stochastic transport principle and on a decomposition of the fluid flow into a largescale component and a noise term that models the unresolved flow components. As for the classical deterministic system, this scheme, referred to as modelling under location uncertainty LU, conserves the global energy of any realization and provides the possibility to generate an ensemble of physically relevant random simulations with a good tradeoff between the model error representation and the ensemble's spread. To maintain numerically the energy conservation feature, we combine an energy in space preserving discretization of the underlying deterministic model with approximations of the stochastic terms that are based on standard finite volumedifference operators. The LU derivation, built from the very same conservation principles as the usual geophysical models, together with the numerical scheme proposed can be directly used in existing dynamical cores of global numerical weather prediction models. The capabilities of the proposed framework is demonstrated for an inviscid test case on the fplane and for a barotropically unstable jet on the sphere.

AttDMM An Attentive Deep Markov Model for Risk Scoring in Intensive Care Units ; Clinical practice in intensive care units ICUs requires early warnings when a patient's condition is about to deteriorate so that preventive measures can be undertaken. To this end, prediction algorithms have been developed that estimate the risk of mortality in ICUs. In this work, we propose a novel generative deep probabilistic model for realtime risk scoring in ICUs. Specifically, we develop an attentive deep Markov model called AttDMM. To the best of our knowledge, AttDMM is the first ICU prediction model that jointly learns both longterm disease dynamics via attention and different disease states in health trajectory via a latent variable model. Our evaluations were based on an established baseline dataset MIMICIII with 53,423 ICU stays. The results confirm that compared to stateoftheart baselines, our AttDMM was superior AttDMM achieved an area under the receiver operating characteristic curve AUROC of 0.876, which yielded an improvement over the stateoftheart method by 2.2. In addition, the risk score from the AttDMM provided warnings several hours earlier. Thereby, our model shows a path towards identifying patients at risk so that health practitioners can intervene early and save patient lives.

Player Modeling via MultiArmed Bandits ; This paper focuses on building personalized player models solely from player behavior in the context of adaptive games. We present two main contributions The first is a novel approach to player modeling based on multiarmed bandits MABs. This approach addresses, at the same time and in a principled way, both the problem of collecting data to model the characteristics of interest for the current player and the problem of adapting the interactive experience based on this model. Second, we present an approach to evaluating and finetuning these algorithms prior to generating data in a user study. This is an important problem, because conducting user studies is an expensive and laborintensive process; therefore, an ability to evaluate the algorithms beforehand can save a significant amount of resources. We evaluate our approach in the context of modeling players' social comparison orientation SCO and present empirical results from both simulations and real players.

Selfdual models in D21 from dimensional reduction ; Here we perform a KaluzaKlein dimensional reduction of Vasiliev's firstorder description of massless spins particles from D31 to D21 and derive firstorder selfdual models describing particles with helicities pm s for the cases s1,2,3. In the first two cases we recover known parity singlets selfdual models. In the spin3 case we derive a new first order selfdual model with a local Weyl symmetry which lifts the traceless restriction on the rank3 tensor. A gauge fixed version of this model corresponds to a known spin3 selfdual model. We conjecture that our procedure can be generalized to arbitrary integer spins.

Improving ModelBased Reinforcement Learning with Internal State Representations through SelfSupervision ; Using a model of the environment, reinforcement learning agents can plan their future moves and achieve superhuman performance in board games like Chess, Shogi, and Go, while remaining relatively sampleefficient. As demonstrated by the MuZero Algorithm, the environment model can even be learned dynamically, generalizing the agent to many more tasks while at the same time achieving stateoftheart performance. Notably, MuZero uses internal state representations derived from real environment states for its predictions. In this paper, we bind the model's predicted internal state representation to the environment state via two additional terms a reconstruction model loss and a simpler consistency loss, both of which work independently and unsupervised, acting as constraints to stabilize the learning process. Our experiments show that this new integration of reconstruction model loss and simpler consistency loss provide a significant performance increase in OpenAI Gym environments. Our modifications also enable selfsupervised pretraining for MuZero, so the algorithm can learn about environment dynamics before a goal is made available.

Deep learning model for multiwavelength emission from lowluminosity active galactic nuclei ; Most active supermassive black holes SMBH in presentday galaxies are underfed and consist of lowluminosity active galactic nuclei LLAGN. They have multiwavelength broadband spectral energy distributions SED dominated by nonthermal processes which are quite different from those of the brighter, more distant quasars. Modelling the observed SEDs of LLAGNs is currently challenging, given the large computational expenses required. In this work, we used machine learning ML methods to generate model SEDs and fit sparse observations of LLAGNs. Our ML model consisted of a neural network and reproduced with excellent precision the radiotoXrays emission from a radiatively inefficient accretion flow around a SMBH and a relativistic jet, at a small fraction of the computational cost. The ML method performs the fit 4 times 105 times faster than previous semianalytic models. As a proofofconcept, we used the ML model to reproduce the SEDs of the LLAGNs M87, NGC 315 and NGC 4261.

The effective model structure and inftygroupoid objects ; For a category mathcal E with finite limits and wellbehaved countable coproducts, we construct a model structure, called the effective model structure, on the category of simplicial objects in mathcal E, generalising the KanQuillen model structure on simplicial sets. We then prove that the effective model structure is left and right proper and satisfies descent in the sense of Rezk. As a consequence, we obtain that the associated inftycategory has finite limits, colimits satisfying descent, and is locally Cartesian closed when mathcal E is, but is not a higher topos in general. We also characterise the inftycategory presented by the effective model structure, showing that it is the full subcategory of presheaves on mathcal E spanned by Kan complexes in mathcal E, a result that suggests a close analogy with the theory of exact completions.

Modelfree Representation Learning and Exploration in Lowrank MDPs ; The low rank MDP has emerged as an important model for studying representation learning and exploration in reinforcement learning. With a known representation, several modelfree exploration strategies exist. In contrast, all algorithms for the unknown representation setting are modelbased, thereby requiring the ability to model the full dynamics. In this work, we present the first modelfree representation learning algorithms for low rank MDPs. The key algorithmic contribution is a new minimax representation learning objective, for which we provide variants with differing tradeoffs in their statistical and computational properties. We interleave this representation learning step with an exploration strategy to cover the state space in a rewardfree manner. The resulting algorithms are provably sample efficient and can accommodate general function approximation to scale to complex environments.

Modelbased Meta Reinforcement Learning using Graph Structured Surrogate Models ; Reinforcement learning is a promising paradigm for solving sequential decisionmaking problems, but low data efficiency and weak generalization across tasks are bottlenecks in realworld applications. Modelbased meta reinforcement learning addresses these issues by learning dynamics and leveraging knowledge from prior experience. In this paper, we take a closer look at this framework, and propose a new Thompsonsampling based approach that consists of a new model to identify task dynamics together with an amortized policy optimization step. We show that our model, called a graph structured surrogate model GSSM, outperforms stateoftheart methods in predicting environment dynamics. Additionally, our approach is able to obtain high returns, while allowing fast execution during deployment by avoiding test time policy gradient optimization.

Improved Denoising Diffusion Probabilistic Models ; Denoising diffusion probabilistic models DDPM are a class of generative models which have recently been shown to produce excellent samples. We show that with a few simple modifications, DDPMs can also achieve competitive loglikelihoods while maintaining high sample quality. Additionally, we find that learning variances of the reverse diffusion process allows sampling with an order of magnitude fewer forward passes with a negligible difference in sample quality, which is important for the practical deployment of these models. We additionally use precision and recall to compare how well DDPMs and GANs cover the target distribution. Finally, we show that the sample quality and likelihood of these models scale smoothly with model capacity and training compute, making them easily scalable. We release our code at httpsgithub.comopenaiimproveddiffusion

An Empirical Study on Measuring the Similarity of Sentential Arguments with Language Model Domain Adaptation ; Measuring the similarity between two different sentential arguments is an important task in argument mining. However, one of the challenges in this field is that the dataset must be annotated using expertise in a variety of topics, making supervised learning with labeled data expensive. In this paper, we investigated whether this problem could be alleviated through transfer learning. We first adapted a pretrained language model to a domain of interest using selfsupervised learning. Then, we finetuned the model to a task of measuring the similarity between sentences taken from different domains. Our approach improves a correlation with humanannotated similarity scores compared to competitive baseline models on the Argument Facet Similarity dataset in an unsupervised setting. Moreover, we achieve comparable performance to a fully supervised baseline model by using only about 60 of the labeled data samples. We believe that our work suggests the possibility of a generalized argument clustering model for various argumentative topics.

JSTRR Model Joint Modeling of Ratings and Reviews in SentimentTopic Prediction ; Analysis of online reviews has attracted great attention with broad applications. Often times, the textual reviews are coupled with the numerical ratings in the data. In this work, we propose a probabilistic model to accommodate both textual reviews and overall ratings with consideration of their intrinsic connection for a joint sentimenttopic prediction. The key of the proposed method is to develop a unified generative model where the topic modeling is constructed based on review texts and the sentiment prediction is obtained by combining review texts and overall ratings. The inference of model parameters are obtained by an efficient Gibbs sampling procedure. The proposed method can enhance the prediction accuracy of review data and achieve an effective detection of interpretable topics and sentiments. The merits of the proposed method are elaborated by the case study from Amazon datasets and simulation studies.

An isotropic compact stellar model in curvature coordinate system consistent with observational data ; This paper investigates a spherically symmetric compact relativistic body with isotropic pressure profiles within the framework of general relativity. In order to solve the Einstein's field equations, we have considered the VaidyaTikekar type metric potential, which depends upon parameter K. We have presented a perfect fluid model, considering K0 or K1, which represent compact stars like SMC X1, Her X1, 4U 153852, SAX J1808.43658, LMC X4, EXO 1785248 and 4U182030, to an excellent degree of accuracy. We have investigated the physical features such as the energy conditions, velocity of sound, surface redshift, adiabatic index of the model in detail and shown that our model obeys all the physical requirements for a realistic stellar model. Using the TolmanOppenheimerVolkoff equations, we have explored the hydrostatic equilibrium and the stability of the compact objects. This model also fulfils the HarrisonZeldovichNovikov stability criterion. The results obtained in this paper can be used in analyzing other isotropic compact objects.

Doing Good or Doing Right Exploring the Weakness of Commonsense Causal Reasoning Models ; Pretrained language models PLM achieve surprising performance on the Choice of Plausible Alternatives COPA task. However, whether PLMs have truly acquired the ability of causal reasoning remains a question. In this paper, we investigate the problem of semantic similarity bias and reveal the vulnerability of current COPA models by certain attacks. Previous solutions that tackle the superficial cues of unbalanced token distribution still encounter the same problem of semantic bias, even more seriously due to the utilization of more training data. We mitigate this problem by simply adding a regularization loss and experimental results show that this solution not only improves the model's generalization ability, but also assists the models to perform more robustly on a challenging dataset, BCOPACE, which has unbiased token distribution and is more difficult for models to distinguish cause and effect.

Financial Markets and the Phase Transition between Water and Steam ; Motivated by empirical observations on the interplay of trends and reversion, a lattice gas model of financial markets is presented. The shares of an asset are modeled by gas molecules that are distributed across a hidden social network of investors. The model is equivalent to the Ising model on this network, whose magnetization represents the deviation of the asset price from its value. Moreover, the system should drive itself to its critical temperature in efficient markets. There, it is characterized by universal critical exponents, in analogy with the secondorder phase transition between water and steam. These critical exponents imply predictions for the autocorrelations of financial market returns and for Hurst exponents. For a simple network topology, consistency with empirical observations implies a fractal network dimension near 3, and a correlation time at least as long as the economic cyle. To also explain the observed market autocorrelations at intermediate scales, the model should be extended beyond the critical domain, to other network topologies, and to other models of critical dynamics.

Redshift drift in radially inhomogeneous LemaitreTolmanBondi spacetimes ; We provide a formula for estimating the redshift and its secular change redshift drift in LemaitreTolmanBondi LTB spherically symmetric universes. We compute the scaling of the redshift drift for LTB models that predict Hubble diagrams indistinguishable from those of the standard cosmological model, the flat Lambda Cold Dark Matter LambdaCDM model. We show that the redshift drift for these degenerate LTB models is typically different from that predicted in the LambdaCDM scenario. We also highlight and discuss some unconventional redshiftdrift signals that arise in LTB universes and give them distinctive features compared to the standard model. We argue that the redshift drift is a metric observable that allows to reduce the degrees of freedom of spherically symmetric models and to make them more predictive and thus falsifiable.

From Common Sense Reasoning to Neural Network Models through Multiple Preferences an overview ; In this paper we discuss the relationships between conditional and preferential logics and neural network models, based on a multipreferential semantics. We propose a conceptwise multipreference semantics, recently introduced for defeasible description logics to take into account preferences with respect to different concepts, as a tool for providing a semantic interpretation to neural network models. This approach has been explored both for unsupervised neural network models SelfOrganising Maps and for supervised ones Multilayer Perceptrons, and we expect that the same approach might be extended to other neural network models. It allows for logical properties of the network to be checked by model checking over an interpretation capturing the inputoutput behavior of the network. For Multilayer Perceptrons, the deep network itself can be regarded as a conditional knowledge base, in which synaptic connections correspond to weighted conditionals. The paper describes the general approach, through the cases of SelfOrganising Maps and Multilayer Perceptrons, and discusses some open issues and perspectives.

On a kinetic opinion formation model for preelection polling ; Motivated by recent successes in modelbased preelection polling, we propose a kinetic model for opinion formation which includes voter demographics and socioeconomic factors like age, sex, ethnicity, education level, income and other measurable factors like behaviour in previous elections or referenda as a key driver in the opinion formation dynamics. The model is based on Toscani's kinetic opinion formation model and the leaderfollower model of During et al., and leads to a system of coupled Boltzmanntype equations and associated, approximate FokkerPlancktype systems. Numerical examples using data from general elections in the United Kingdom show the effect different demographics have on the opinion formation process and the outcome of elections.

Beyond InPlace Corruption Insertion and Deletion In Denoising Probabilistic Models ; Denoising diffusion probabilistic models DDPMs have shown impressive results on sequence generation by iteratively corrupting each example and then learning to map corrupted versions back to the original. However, previous work has largely focused on inplace corruption, adding noise to each pixel or token individually while keeping their locations the same. In this work, we consider a broader class of corruption processes and denoising models over sequence data that can insert and delete elements, while still being efficient to train and sample from. We demonstrate that these models outperform standard inplace models on an arithmetic sequence task, and that when trained on the text8 dataset they can be used to fix spelling errors without any finetuning.

A Fluctuating LineofSight Fading Model with DoubleRayleigh Diffuse Scattering ; We introduce the fdRLoS fading model as a natural generalization of the doubleRayleigh with lineofsight fading model, on which the constantamplitude lineofsight component is now allowed to randomly fluctuate. We discuss the key benefits of the fdRLoS fading model here formulated over the state of the art, and provide an analytical characterization of its chief probability functions. We analyze the effect of the fading parameters that define the model, and discuss their impact on the performance of wireless communication systems.

Towards a Higgs mass determination in asymptotically safe gravity with a dark portal ; There are indications that an asymptotically safe UV completion of the Standard Model with gravity could constrain the Higgs selfcoupling, resulting in a prediction of the Higgs mass close to the vacuum stability bound in the Standard Model. The predicted value depends on the top quark mass and comes out somewhat higher than the experimental value if the current central value for the top quark mass is assumed. Beyond the Standard Model, the predicted value also depends on dark fields coupled through a Higgs portal. Here we study the Higgs selfcoupling in a toy model of the Standard Model with quantum gravity that we extend by a dark scalar and fermion. Within the approximations used in arXiv2005.03661 , there is a single free parameter in the asymptotically safe dark sector, as a function of which the predicted toy model Higgs mass can be lowered due to mixing effects if the dark sector undergoes spontaneous symmetry breaking.

Audio Captioning Transformer ; Audio captioning aims to automatically generate a natural language description of an audio clip. Most captioning models follow an encoderdecoder architecture, where the decoder predicts words based on the audio features extracted by the encoder. Convolutional neural networks CNNs and recurrent neural networks RNNs are often used as the audio encoder. However, CNNs can be limited in modelling temporal relationships among the time frames in an audio signal, while RNNs can be limited in modelling the longrange dependencies among the time frames. In this paper, we propose an Audio Captioning Transformer ACT, which is a full Transformer network based on an encoderdecoder architecture and is totally convolutionfree. The proposed method has a better ability to model the global information within an audio signal as well as capture temporal relationships between audio events. We evaluate our model on AudioCaps, which is the largest audio captioning dataset publicly available. Our model shows competitive performance compared to other stateoftheart approaches.

Unsupervised Detection of Adversarial Examples with Model Explanations ; Deep Neural Networks DNNs have shown remarkable performance in a diverse range of machine learning applications. However, it is widely known that DNNs are vulnerable to simple adversarial perturbations, which causes the model to incorrectly classify inputs. In this paper, we propose a simple yet effective method to detect adversarial examples, using methods developed to explain the model's behavior. Our key observation is that adding small, humanly imperceptible perturbations can lead to drastic changes in the model explanations, resulting in unusual or irregular forms of explanations. From this insight, we propose an unsupervised detection of adversarial examples using reconstructor networks trained only on model explanations of benign examples. Our evaluations with MNIST handwritten dataset show that our method is capable of detecting adversarial examples generated by the stateoftheart algorithms with high confidence. To the best of our knowledge, this work is the first in suggesting unsupervised defense method using model explanations.

Identifying the fragment structure of the organic compounds by deeply learning the original NMR data ; We preprocess the raw NMR spectrum and extract key characteristic features by using two different methodologies, called equidistant sampling and peak sampling for subsequent substructure pattern recognition; meanwhile may provide the alternative strategy to address the imbalance issue of the NMR dataset frequently encountered in dataset collection of statistical modeling and establish two conventional SVM and KNN models to assess the capability of two feature selection, respectively. Our results in this study show that the models using the selected features of peak sampling outperform the ones using the other. Then we build the Recurrent Neural Network RNN model trained by Data B collected from peak sampling. Furthermore, we illustrate the easier optimization of hyper parameters and the better generalization ability of the RNN deep learning model by comparison with traditional machine learning SVM and KNN models in detail.

Can we infer player behavior tendencies from a player's decisionmaking data Integrating Theory of Mind to Player Modeling ; Game AI systems need the theory of mind, which is the humanistic ability to infer others' mental models, preferences, and intent. Such systems would enable inferring players' behavior tendencies that contribute to the variations in their decisionmaking behaviors. To that end, in this paper, we propose the use of inverse Bayesian inference to infer behavior tendencies given a descriptive cognitive model of a player's decision making. The model embeds behavior tendencies as weight parameters in a player's decisionmaking. Inferences on such parameters provide intuitive interpretations about a player's cognition while making ingame decisions. We illustrate the use of inverse Bayesian inference with synthetically generated data in a game called textitBoomTown developed by Gallup. We use the proposed model to infer a player's behavior tendencies for moving decisions on a game map. Our results indicate that our model is able to infer these parameters towards uncovering not only a player's decision making but also their behavior tendencies for making such decisions.

Model parameters in the context of late time cosmic acceleration in fQ,T gravity ; The dynamical aspects of some accelerating models are investigated in the framework of an extension of symmetric teleparllel gravity dubbed as fQ,T gravity. In this gravity theory, the usual Ricci tensor in the geometrical action is replaced by a functional fQ,T where Q is the nonmetricity and T is the trace of the energymomentum tensor. Two different functional forms are considered in the present work. In order to model the Universe, we have considered a signature flipping deceleration parameter simulated by a hybrid scale factor HSF. The dynamical parameters of the model are derived and analysed. We discuss the role of the parameter space in getting viable cosmological models. It is found that, the models may be useful as suitable geometrical alternatives to the usual dark energy approach.

SelfSupervised Inference in StateSpace Models ; We perform approximate inference in statespace models with nonlinear state transitions. Without parameterizing a generative model, we apply Bayesian update formulas using a local linearity approximation parameterized by neural networks. This comes accompanied by a maximum likelihood objective that requires no supervision via uncorrupt observations or ground truth latent states. The optimization backpropagates through a recursion similar to the classical Kalman filter and smoother. Additionally, using an approximate conditional independence, we can perform smoothing without having to parameterize a separate model. In scientific applications, domain knowledge can give a linear approximation of the latent transition maps, which we can easily incorporate into our model. Usage of such domain knowledge is reflected in excellent results despite our model's simplicity on the chaotic Lorenz system compared to fully supervised and variational inference methods. Finally, we show competitive results on an audio denoising experiment.

Renewable Energy Targets and Unintended Storage Cycling Implications for Energy Modeling ; To decarbonize the economy, many governments have set targets for the use of renewable energy sources. These are often formulated as relative shares of electricity demand or supply. Implementing respective constraints in energy models is a surprisingly delicate issue. They may cause a modeling artifact of excessive electricity storage use. We introduce this phenomenon as 'unintended storage cycling', which can be detected in case of simultaneous storage charging and discharging. In this paper, we provide an analytical representation of different approaches for implementing minimum renewable share constraints in models, and show how these may lead to unintended storage cycling. Using a parsimonious optimization model, we quantify related distortions of optimal dispatch and investment decisions as well as market prices, and identify important drivers of the phenomenon. Finally, we provide recommendations on how to avoid the distorting effects of unintended storage cycling in energy modeling.

Constraining the dark energy models using Baryon Acoustic Oscillations An approach independent of H0 cdot rd ; The H0 tension and the accompanying rd tension are a hot topic in current cosmology. In order to remove the degeneracy between the Hubble parameter H0 and the sound horizon scale rd from the Baryon Acoustic Oscillations BAO datasets, we redefine the likelihood by marginalizing over the H0 cdot rd parameter and then we perform full Bayesian analysis for different models of dark energy DE. We find that our uncalibrated by early or late physics datasets cannot constrain the DE models properly without further assumptions. By adding the type IA supernova dataset, the models are constrained better with smaller errors on the DE parameters. The two BAO datasets we use one with angular measurements and one with angular and radial ones with their covariances, show statistical preferences for different models, with LambdaCDM being the best model for one of them. Adding the Pantheon SnIA dataset with its covariance matrix boosts the statistical preference for LambdaCDM.

Shallow shell models by Gamma convergence ; In this paper we derive, by means of Gammaconvergence, the shallow shell models starting from non linear three dimensional elasticity. We use the approach analogous to the one for shells and plates. We start from the minimization formulation of the general three dimensional elastic body which is subjected to normal volume forces and free boundary conditions and do not presuppose any constitutional behavior. To derive the model we need to propose how is the order of magnitudes of the external loads related to the thickness of the body h as well as the order of the geometry of the shallow shell. We analyze the situation when the external normal forces are of order halpha, where alpha2. For alpha3 we obtain the Marguerrevon K'arm'an model and for alpha3 the linearized Marguerrevon K'arm'an model. For alpha in 2,3 we are able to obtain only the lower bound for the Gammalimit. This is analogous to the recent results for the ordinary shell models.

A flowing plasma model to describe drift waves in a cylindrical helicon discharge ; A twofluid model developed originally to describe wave oscillations in the vacuum arc centrifuge, a cylindrical, rapidly rotating, low temperature and confined plasma column, is applied to interpret plasma oscillations in a RF generated linear magnetised plasma WOMBAT, with similar density and field strength. Compared to typical centrifuge plasmas, WOMBAT plasmas have slower normalised rotation frequency, lower temperature and lower axial velocity. Despite these differences, the twofluid model provides a consistent description of the WOMBAT plasma configuration and yields qualitative agreement between measured and predicted wave oscillation frequencies with axial field strength. In addition, the radial profile of the density perturbation predicted by this model is consistent with the data. Parameter scans show that the dispersion curve is sensitive to the axial field strength and the electron temperature, and the dependence of oscillation frequency with electron temperature matches the experiment. These results consolidate earlier claims that the density and floating potential oscillations are a resistive drift mode, driven by the density gradient. To our knowledge, this is the first detailed physics model of flowing plasmas in the diffusion region away from the RF source. Possible extensions to the model, including temperature nonuniformity and magnetic field oscillations, are also discussed.

Evolutionary Dynamics in a Simple Model of SelfAssembly ; We investigate the evolutionary dynamics of an idealised model for the robust selfassembly of twodimensional structures called polyominoes. The model includes rules that encode interactions between sets of square tiles that drive the selfassembly process. The relationship between the model's rule set and its resulting selfassembled structure can be viewed as a genotypephenotype map and incorporated into a genetic algorithm. The rule sets evolve under selection for specified target structures. The corresponding, complex fitness landscape generates rich evolutionary dynamics as a function of parameters such as the population size, search space size, mutation rate, and method of recombination. Furthermore, these systems are simple enough that in some cases the associated model genome space can be completely characterised, shedding light on how the evolutionary dynamics depends on the detailed structure of the fitness landscape. Finally, we apply the model to study the emergence of the preference for dihedral over cyclic symmetry observed for homomeric protein tetramers.

A Comprehensive Trainable Error Model for Sung Music Queries ; We propose a model for errors in sung queries, a variant of the hidden Markov model HMM. This is a solution to the problem of identifying the degree of similarity between a typically errorladen sung query and a potential target in a database of musical works, an important problem in the field of music information retrieval. Similarity metrics are a critical component of querybyhumming QBH applications which search audio and multimedia databases for strong matches to oral queries. Our model comprehensively expresses the types of error or variation between target and query cumulative and noncumulative local errors, transposition, tempo and tempo changes, insertions, deletions and modulation. The model is not only expressive, but automatically trainable, or able to learn and generalize from query examples. We present results of simulations, designed to assess the discriminatory potential of the model, and tests with real sung queries, to demonstrate relevance to realworld applications.

TopBESS model and its phenomenology ; We introduce the topBESS model which is the effective description of the strong electroweak symmetry breaking with a single new SU2LR triplet vector resonance. The model is a modification of the BESS model in the fermion sector. The triplet couples to the third generation of quarks only. This approach reflects a possible extraordinary role of the top quark in the mechanism of electroweak symmetry breaking. The lowenergy limits on the model parameters found provide hope for finding sizable signals in the LHC DrellYan processes as well as in the schannel production processes at the ILC. However, there are regions of the model parameter space where the interplay of the direct and indirect fermion couplings can hide the resonance peak in a scattering process even though the resonance exists and couples directly to top and bottom quarks.

Linear Sigma Models with Torsion ; Gauged linear sigma models with 0,2 supersymmetry allow a larger choice of couplings than models with 2,2 supersymmetry. We use this freedom to find a fully linear construction of torsional heterotic compactifications, including models with branes. As a noncompact example, we describe a family of metrics which correspond to deformations of the heterotic conifold by turning on Hflux. We then describe compact models which are gaugeinvariant only at the quantum level. Our construction gives a generalization of symplectic reduction. The resulting spaces are nonKahler analogues of familiar toric spaces like complex projective space. Perturbatively conformal models can be constructed by considering intersections.

The critical exponents of the QCD tricritical endpoint within exactly solvable models ; The critical indices alpha', beta, gamma' and delta of the Quark Gluon Bags with Surface Tension Model with the tricritical and critical endpoint are calculated as functions of the usual parameters of this model and two newly introduced parameters indices. The critical indices are compared with that ones of other models. The universality class of the present model with respect to values of the model parameters is discussed. The scaling relations for the found critical exponents are verified and it is demonstrated that for the standard definition of the index alpha' some of them are not fulfilled in general case. Although it is shown that the specially defined index alpha's recovers the scaling relations, another possibility, an existence of the nonFisher universality classes, is also discussed.

Search for Contact Interactions in epmp Collisions at HERA ; A search for physics beyond the Standard Model in neutral current deep inelastic scattering at high negative fourmomentum transfer squared Q2 is performed in epmp collisions at HERA. The differential cross section dsdQ2, measured using the full H1 data sample corresponding to an integrated luminosity of 446 pb1, is compared to the Standard Model prediction. No significant deviation is observed. Limits on various models predicting new phenomena at high Q2 are derived. For general fourfermion eeqq contact interaction models, lower limits on the compositeness scale Lambda are set in the range 3.6 TeV to 7.2 TeV. Leptoquarks with masses MLQ and couplings lambda are constrained to MLQlambda 0.411.86 TeV and limits on squarks in Rparity violating supersymmetric models are derived. A lower limit on the gravitational scale in 4n dimensions of MS 0.9 TeV is established for lowscale quantum gravity effects in models with large extra dimensions. For the light quark radius an upper bound of Rq 0.65 1018 m is determined.

Interference minimization in physical model of wireless networks ; Interference minimization problem in wireless sensor and adhoc networks is considered. That is to assign a transmission power to each node of a network such that the network is connected and at the same time the maximum of accumulated signal straight on network nodes is minimum. Previous works on interference minimization in wireless networks mainly consider the disk graph model of network. For disk graph model two approximation algorithms with Osqrtn and Ooptlnn2 upper bounds of maximum interference are known, where n is the number of nodes and opt is the minimal interference of a given network. In current work we consider more general interference model, the physical interference model, where sender nodes' signal straight on a given node is a function of a senderreceiver node pair and sender nodes' transmission power. For this model we give a polynomial time approximation algorithm which finds a connected network with at most Ooptlnn2beta interference, where beta geq 1 is the minimum signal straight necessary on receiver node for successfully receiving a message.

A firstprinciples model of timedependent variations in transmission through a fluctuating scattering environment ; Fading is the timedependent variation in transmitted signal strength through a complex medium, due to interference or temporally evolving multipath scattering. In this paper we use random matrix theory RMT to establish a firstprinciples model for fading, including both universal and nonuniversal effects. This model provides a more general understanding of the most common statistical models Rayleigh fading and Rice fading and provides a detailed physical basis for their parameters. We also report experimental tests on two raychaotic microwave cavities. The results show that our RMT model agrees with the RayleighRice models in the high loss regime, but there are strong deviations in lowloss systems where the RMT approach describes the data well.

Smooth Hamiltonian systems with soft impacts ; In a Hamiltonian system with impacts or billiard with potential, a point particle moves about the interior of a bounded domain according to a background potential, and undergoes elastic collisions at the boundaries. When the background potential is identically zero, this is the hardwall billiard model. Previous results on smooth billiard models where the hardwall boundary is replaced by a steep smooth billiardlike potential have clarified how the approximation of a smooth billiard with a hardwall billiard may be utilized rigorously. These results are extended here to models with smooth background potential satisfying some natural conditions. This generalization is then applied to geometric models of collinear triatomic chemical reactions the models are far from integrable ndegree of freedom systems with ngeq2. The application demonstrates that the simpler analytical calculations for the hardwall system may be used to obtain qualitative information with regard to the solution structure of the smooth system and to quantitatively assist in finding solutions of the soft impact system by continuation methods. In particular, stable periodic triatomic configurations are easily located for the smooth highlynonlinear two and three degree of freedom geometric models.

Extremal Aging For Trap Models ; In the seminal work 5, Ben Arous and vCern'y give a general characterization of aging for trap models in terms of alphastable subordinators with alpha in 0,1. Some of the important examples that fall into this universality class are Random Hopping Time RHT dynamics of Random Energy Model REM and pspin models observed on exponential time scales. In this paper, we explain a different aging mechanism in terms of it extremal processes that can be seen as the extension of alphastable aging to the case alpha0. We apply this mechanism to the RHT dynamics of the REM for a wide range of temperature and time scales. The other examples that exhibit extremal aging include the Sherrington Kirkpatrick SK model and pspin models 6, 9, and biased random walk on critical GaltonWatson trees conditioned to survive 11.

Market models with optimal arbitrage ; We construct and study market models admitting optimal arbitrage. We say that a model admits optimal arbitrage if it is possible, in a zerointerest rate setting, starting with an initial wealth of 1 and using only positive portfolios, to superreplicate a constant c1. The optimal arbitrage strategy is the strategy for which this constant has the highest possible value. Our definition of optimal arbitrage is similar to the one in Fernholz and Karatzas 2010, where optimal relative arbitrage with respect to the market portfolio is studied. In this work we present a systematic method to construct market models where the optimal arbitrage strategy exists and is known explicitly. We then develop several new examples of market models with arbitrage, which are based on economic agents' views concerning the impossibility of certain events rather than ad hoc constructions. We also explore the concept of fragility of arbitrage introduced in Guasoni and Rasonyi 2012, and provide new examples of arbitrage models which are not fragile in this sense.

Estimating timechanges in noisy Levy models ; In quantitative finance, we often model asset prices as a noisy Ito semimartingale. As this model is not identifiable, approximating by a timechanged Levy process can be useful for generative modelling. We give a new estimate of the normalised volatility or time change in this model, which obtains minimax convergence rates, and is unaffected by infinitevariation jumps. In the semimartingale model, our estimate remains accurate for the normalised volatility, obtaining convergence rates as good as any previously implied in the literature.

Proceedings of the Fourth International Workshop on DomainSpecific Languages and Models for Robotic Systems DSLRob 2013 ; The Fourth International Workshop on DomainSpecific Languages and Models for Robotic Systems DSLRob'13 was held in conjunction with the 2013 IEEERSJ International Conference on Intelligent Robots and Systems IROS 2013, November 2013 in Tokyo, Japan. The main topics of the workshop were DomainSpecific Languages DSLs and Modeldriven Software Development MDSD for robotics. A domainspecific language is a programming language dedicated to a particular problem domain that offers specific notations and abstractions that increase programmer productivity within that domain. Modeldriven software development offers a highlevel way for domain users to specify the functionality of their system at the right level of abstraction. DSLs and models have historically been used for programming complex systems. However recently they have garnered interest as a separate field of study. Robotic systems blend hardware and software in a holistic way that intrinsically raises many crosscutting concerns concurrency, uncertainty, time constraints, ..., for which reason, traditional generalpurpose languages often lead to a poor fit between the language features and the implementation requirements. DSLs and models offer a powerful, systematic way to overcome this problem, enabling the programmer to quickly and precisely implement novel software solutions to complex problems within the robotics domain.

Gauge Invariance and SpinonDopon Confinement in the tJ Model implications for Fermi Surface Reconstruction in the Cuprates ; We discuss the application of the twoband spindopon representation of the tJ model to address the issue of the Fermi surface reconstruction observed in the cuprates. We show that the electron no double occupancy NDO constraint plays a key role in this formulation. In particular, the auxiliary lattice spin and itinerant dopon degrees of freedom of the spindopon formulation of the tJ model are shown to be confined in the emergent U1 gauge theory generated by the NDO constraint. This constraint is enforced by the requirement of an infinitely large spindopon coupling. As a result, the tJ model is equivalent to a KondoHeisenberg lattice model of itinerant dopons and localized lattice spins at infinite Kondo coupling at all dopings. We show that meanfield treatment of the large vs small Fermi surface crossing in the cuprates which leaves out the NDO constraint, leads to inconsistencies and it is automatically excluded form the t J model framework.

Twodimensional state sum models and spin structures ; The state sum models in two dimensions introduced by Fukuma, Hosono and Kawai are generalised by allowing algebraic data from a nonsymmetric Frobenius algebra. Without any further data, this leads to a state sum model on the sphere. When the data is augmented with a crossing map, the partition function is defined for any oriented surface with a spin structure. An algebraic condition that is necessary for the state sum model to be sensitive to spin structure is determined. Some examples of state sum models that distinguish topologicallyinequivalent spin structures are calculated.

Redshift drift test of exotic singularity universes ; We discuss how dynamical dark energy universes with exotic singularities may be distinguished from the standard LambdaCDM model on the basis of their redshift drift signal, for which measurements both in the acceleration phase and in the deep matter era will be provided by forthcoming astrophysical facilities. Two specific classes of exotic singularity models are studied sudden future singularity models and finite scale factor singularity models. In each class we identify the models which can mimic LambdaCDM and play the role of dark energy as well as models for which redshift drift signals are significantly different from LambdaCDM and the test can differentiate between them.

Neron models of jacobians over base schemes of dimension greater than 1 ; We investigate to what extent the theory of N'eron models of jacobians and of abeljacobi maps extends to relative curves over base schemes of dimension greater than 1. We give a necessary and sufficient criterion for the existence of a N'eron model. We use this to show that, in general, N'eron models do not exist even after making a modification or even alteration of the base. On the other hand, we show that N'eron models do exist outside some codimension2 locus.

The Deffuant model on mathbbZ with higherdimensional opinion spaces ; When it comes to the mathematical modelling of social interaction patterns, a number of different models have emerged and been studied over the last decade, in which individuals randomly interact on the basis of an underlying graph structure and share their opinions. A prominent example of the socalled bounded confidence models is the one introduced by Deffuant et al. Two neighboring individuals will only interact if their opinions do not differ by more than a given threshold theta. We consider this model on the line graph mathbbZ and extend the results that have been achieved for the model with realvalued opinions by considering vectorvalued opinions and general metrics measuring the distance between two opinion values. Just as in the univariate case, there exists a critical value for theta at which a phase transition in the longterm behavior takes place.

Higher order corrections of the extended Chaplygin gas cosmology with varying G and ; In this paper, we study two different models of dark energy based on Chaplygin gas equation of state. The first model is the variable modified Chaplygin gas while the second one is the extended Chaplygin gas. Both models are considered in the framework of higher order fR modified gravity. We also consider the case of time varying gravitational constant G and Lambda for both models. We investigate some cosmological parameters such as the Hubble, the deceleration and the equation of state parameters. Then we showed that the model that we considered, extended Chaplygin gas with timedependent G and Lambda, is consistent with the observational data. Finally we conclude with the discussion of cosmological perturbations of our model.

Effect of an external interaction mechanism in solving agegraphic dark energy problems ; Agegraphic dark energyADE and NewADE models have been introduced as two candidates for dark energy to explain the accelerated expansion phase of the Universe. In spite of a few suitable features of these models some studies have shown that there are several drawbacks in them. Therefore in this investigation a new version of ADE and NewADE are studied which can improve such drawbacks which appear in the ordinary ADE and NewADE scenario. In fact we consider an interacting model of scalar field with matter and after rederiving some cosmological parameters of the model, we find out the best fit for the model. Actually by finding the best fitting for free parameters of the model, we show that our theoretical results are in a good agreement with observational data.

Active biopolymer networks generate scalefree but euclidean clusters ; We report analytical and numerical modelling of active elastic networks, motivated by experiments on crosslinked actin networks contracted by myosin motors. Within a broad range of parameters, the motordriven collapse of active elastic networks leads to a critical state. We show that this state is qualitatively different from that of the random percolation model. Intriguingly, it possesses both euclidean and scalefree structure with Fisher exponent smaller than 2. Remarkably, an indistinguishable Fisher exponent and the same euclidean structure is obtained at the critical point of the random percolation model after absorbing all enclaves into their surrounding clusters. We propose that in the experiment the enclaves are absorbed due to steric interactions of network elements. We model the network collapse, taking into account the steric interactions. The model shows how the system robustly drives itself towards the critical point of the random percolation model with absorbed enclaves, in agreement with the experiment.

Fluctuations analysis in complex networks modeled by hidden variable models. Necessity of a large cutoff in hiddenvariable models ; It is becoming more and more clear that complex networks present remarkable large fluctuations. These fluctuations may manifest differently according to the given model. In this paper we reconsider hidden variable models which turn out to be more analytically treatable and for which we have recently shown clear evidence of nonself averaging; the density of a motif being subject to possible uncontrollable fluctuations in the infinite size limit. Here we provide full detailed calculations and we show that large fluctuations are only due to the node hidden variables variability while, in ensembles where these are frozen, fluctuations are negligible in the thermodynamic limit, and equal the fluctuations of classical random graphs. A special attention is paid to the choice of the cutoff we show that in hiddenvariable models, only a cutoff growing as Nlambda with lambdageq 1 can reproduce the scaling of a powerlaw degree distribution. In turn, it is this large cutoff that generates nonselfaveraging.

Characterizing and computing stable models of logic programs The nonstratified case ; Stable Logic Programming SLP is an emergent, alternative style of logic programming each solution to a problem is represented by a stable model of a deductive databasefunctionfree logic program encoding the problem itself. Several implementations now exist for stable logic programming, and their performance is rapidly improving. To make SLP generally applicable, it should be possible to check for consistency i.e., existence of stable models of the input program before attempting to answer queries. In the literature, only rather strong sufficient conditions have been proposed for consistency, e.g., stratification. This paper extends these results in several directions. First, the syntactic features of programs, viz. cyclic negative dependencies, affecting the existence of stable models are characterized, and their relevance is discussed. Next, a new graph representation of logic programs, the Extended Dependency Graph EDG, is introduced, which conveys enough information for reasoning about stable models while the traditional Dependency Graph does not. Finally, we show that the problem of the existence of stable models can be reformulated in terms of coloring of the EDG.

Revisiting the RMDM Models ; Combining neutrino mass generation and a dark matter candidate in a unified model has always been intriguing. We revisit the class of RnuMDM models, which incorporate minimal dark matter in radiative neutrino mass models based on the oneloop ultraviolet completions of the Weinberg operator. The possibility of an exact accidental Z2 is completely ruled out in this scenario. We study the phenomenology of one of the models with an approximate Z2 symmetry. In addition to the Standard Model particles, it contains two real scalar quintuplets, one vector like quadruplet fermion and a fermionic quintuplet. The neutral component of the fermionic quintuplet serves as a good dark matter candidate which can be tested by the future direct and indirect detection experiments. The constraints from flavor physics and electroweakscale naturalness are also discussed.

A Bayesian Model of Multilingual Unsupervised Semantic Role Induction ; We propose a Bayesian model of unsupervised semantic role induction in multiple languages, and use it to explore the usefulness of parallel corpora for this task. Our joint Bayesian model consists of individual models for each language plus additional latent variables that capture alignments between roles across languages. Because it is a generative Bayesian model, we can do evaluations in a variety of scenarios just by varying the inference procedure, without changing the model, thereby comparing the scenarios directly. We compare using only monolingual data, using a parallel corpus, using a parallel corpus with annotations in the other language, and using small amounts of annotation in the target language. We find that the biggest impact of adding a parallel corpus to training is actually the increase in monolingual data, with the alignments to another language resulting in small improvements, even with labeled data for the other language.

Effects of anisotropy on interacting ghost dark energy in BransDicke theories ; By interacting ghost dark energy ghost DE in the framework of BransDicke theory, a spatially homogeneous and anisotropic Bianchi type I Universe has been studied. For this purpose, we use the squared sound speed cs2 whose sign determines the stability of the model. As well as we probe observational constraints on the ghost dark energy models as the unification of dark matter and dark energy by using the latest observational data. In order to do so, we focus on observational determinations of the Hubble expansion rate namely, the expansion history Hz. After that we evaluate the evolution of the growth of perturbations in the linear regime for both ghost DE and BransDicke theory and compare the results with standard FRW and LambdaCDM models. We display the effects of the anisotropy on the evolutionary behavior the ghost DE models where the growth rate is higher in this models. Eventually the growth factor for the LambdaCDM Universe will always fall behind the ghost DE models in an anisotropic Universe

Symbolic Reachability Analysis of B through ProB and LTSmin ; We present a symbolic reachability analysis approach for B that can provide a significant speedup over traditional explicit state model checking. The symbolic analysis is implemented by linking ProB to LTSmin, a highperformance language independent model checker. The link is achieved via LTSmin's PINS interface, allowing ProB to benefit from LTSmin's analysis algorithms, while only writing a few hundred lines of gluecode, along with a bridge between ProB and C using ZeroMQ. ProB supports model checking of several formal specification languages such as B, EventB, Z and TLA. Our experiments are based on a wide variety of BMethod and EventB models to demonstrate the efficiency of the new link. Among the tested categories are state space generation and deadlock detection; but action detection and invariant checking are also feasible in principle. In many cases we observe speedups of several orders of magnitude. We also compare the results with other approaches for improving model checking, such as partial order reduction or symmetry reduction. We thus provide a new scalable, symbolic analysis algorithm for the BMethod and EventB, along with a platform to integrate other model checking improvements via LTSmin in the future.

Closing in on minimal dark matter and radiative neutrino masses ; We study oneloop radiative neutrino mass models in which one of the beyondthestandard model fields is either a hyperchargezero fermion quintet minimal dark matter or a hyperchargezero scalar septet. By systematically classifying all possible oneloop such models we identify various processes that render the neutral component of these representations dark matter cosmologically unstable. Thus, our findings show that these scenarios are in general not reconcilable with dark matter stability unless tiny couplings or additional ad hoc symmetries are assumed, in contrast to minimal dark matter models where stability is entirely due to the standard model gauge symmetry. For some variants based on higherorder loops we find that alpha2 reaches a Landau pole at rather low scales, a couple orders of magnitude from the characteristic scale of the model itself. Thus, we argue that some of these variations although consistent with dark matter stability and phenomenological constraints are hard to reconcile with perturbativity criteria.

TreetoSequence Attentional Neural Machine Translation ; Most of the existing Neural Machine Translation NMT models focus on the conversion of sequential data and do not directly use syntactic information. We propose a novel endtoend syntactic NMT model, extending a sequencetosequence model with the sourceside phrase structure. Our model has an attention mechanism that enables the decoder to generate a translated word while softly aligning it with phrases as well as words of the source sentence. Experimental results on the WAT'15 EnglishtoJapanese dataset demonstrate that our proposed model considerably outperforms sequencetosequence attentional NMT models and compares favorably with the stateoftheart treetostring SMT system.

A Flexible Galerkin Scheme for Option Pricing in Levy Models ; One popular approach to option pricing in L'evy models is through solving the related partial integro differential equation PIDE. For the numerical solution of such equations powerful Galerkin methods have been put forward e.g. by Hilber et al. 2013. As in practice large classes of models are maintained simultaneously, flexibility in the driving L'evy model is crucial for the implementation of these powerful tools. In this article we provide such a flexible finite element Galerkin method. To this end we exploit the Fourier representation of the infinitesimal generator, i.e. the related symbol, which is explicitly available for the most relevant L'evy models. Empirical studies for the Merton, NIG and CGMY model confirm the numerical feasibility of the method.

On twists of smooth plane curves ; Given a smooth curve defined over a field k that admits a nonsingular plane model over overlinek, a fixed separable closure of k, it does not necessarily have a nonsingular plane model defined over the field k. We determine under which conditions this happens and we show an example of such phenomenon. Now, even assuming that such a smooth plane model exists, we wonder about the existence of nonsingular plane models over k for its twists. We characterize twists possessing such models and use such characterization to improve, for the particular case of smooth plane curves, the algorithm to compute twists of nonhyperelliptic curves wrote recently down by the third author. We also show an example of a twist not admitting such nonsingular plane model. As a consequence, we get explicit equations for a nontrivial BrauerSeveri surface. Finally, we obtain a theoretical result to compute all the twists of smooth plane curves with cyclic automorphism group having a kmodel whose automorphism group is generated by a diagonal matrix. Some examples are also provided.

A Matrix Model for NonAbelian Quantum Hall States ; We propose a matrix quantum mechanics for a class of nonAbelian quantum Hall states. The model describes electrons which carry an internal SUp spin. The ground states of the matrix model include spinsinglet generalisations of the MooreRead and ReadRezayi states and, in general, lie in a class previously introduced by Blok and Wen. The effective action for these states is a Up ChernSimons theory. We show how the matrix model can be derived from quantisation of the vortices in this ChernSimons theory and how the matrix model ground states can be reconstructed as correlation functions in the boundary WZW model.

Multiplepoint principle with a scalar singlet extension of the Standard Model ; We suggest a scalar singlet extension of the standard model, in which the multiplepoint principle MPP condition of a vanishing Higgs potential at the Planck scale is realized. Although there have been lots of attempts to realize the MPP at the Planck scale, the realization with keeping naturalness is quite difficult. Our model can easily achieve the MPP at the Planck scale without large Higgs mass corrections. It is worth noting that the electroweak symmetry can be radiatively broken in our model. In the naturalness point of view, the singlet scalar mass should be of cal O1,rm TeV or less. We also consider righthanded neutrino extension of the model for neutrino mass generation. The model does not affect the MPP scenario, and might keep the naturalness with the new particle mass scale beyond TeV, thanks to accidental cancellation of Higgs mass corrections.

Robust splitplot designs for model misspecification ; Many existing methods for constructing optimal splitplot designs, such as Doptimal designs, only focus on minimizing the variances and covariances of the estimation for the fitted model. However, the underlying true model is usually complicated and unknown and the fitted model is often misspecified. If there exist significant effects that are not included in the model, then the estimation could be highly biased. Therefore, a good splitplot designs should be able to simultaneously control the variancescovariances and the bias of the estimation. In this paper, we propose a new method for constructing optimal splitplot designs that are robust for model misspecification. We provide a general form of the loss function used for the Doptimal minimax criterion and apply it to searching for robust splitplot designs. To more efficiently construct designs, we develop an algorithm which combines the anneal algorithm and pointexchange algorithm. We modify the update formulas for calculating the determinant and inverse of the updated matrix and apply them to increasing the computing speed for our developed program.

A deep language model for software code ; Existing language models such as ngrams for software code often fail to capture a long context where dependent code elements scatter far apart. In this paper, we propose a novel approach to build a language model for software code to address this particular issue. Our language model, partly inspired by human memory, is built upon the powerful deep learningbased Long Short Term Memory architecture that is capable of learning longterm dependencies which occur frequently in software code. Results from our intrinsic evaluation on a corpus of Java projects have demonstrated the effectiveness of our language model. This work contributes to realizing our vision for DeepSoft, an endtoend, generic deep learningbased framework for modeling software and its development process.

Exponential Family Mixed Membership Models for SoftClustering of Multivariate Data ; For several years, modelbased clustering methods have successfully tackled many of the challenges presented by dataanalysts. However, as the scope of data analysis has evolved, some problems may be beyond the standard mixture model framework. One such problem is when observations in a dataset come from overlapping clusters, whereby different clusters will possess similar parameters for multiple variables. In this setting, mixed membership models, a soft clustering approach whereby observations are not restricted to single cluster membership, have proved to be an effective tool. In this paper, a method for fitting mixed membership models to data generated by a member of an exponential family is outlined. The method is applied to count data obtained from an ultra running competition, and compared with a standard mixture model approach.

Advanced fluid modelling and PICMCC simulations of lowpressure ccrf discharges ; Comparative studies of capacitively coupled radiofrequency discharges in helium and argon at pressures between 10 and 80 Pa are presented applying two different fluid modelling approaches as well as two independently developed particleincellMonte Carlo collision PICMCC codes. The focus is on the analysis of the range of applicability of a recently proposed fluid model including an improved driftdiffusion approximation for the electron component as well as its comparison with fluid modelling results using the classical driftdiffusion approximation and benchmark results obtained by PICMCC simulations. Main features of this time and spacedependent fluid model are given. It is found that the novel approach shows generally quite good agreement with the macroscopic properties derived by the kinetic simulations and is largely able to characterize qualitatively and quantitatively the discharge behaviour even at conditions when the classical fluid modelling approach fails. Furthermore, the excellent agreement between the two PICMCC simulation codes using the velocity Verlet method for the integration of the equations of motion verifies their accuracy and applicability.

Mathematical model for pestinsect control using mating disruption and trapping ; Controlling pest insects is a challenge of main importance to preserve crop production. In the context of Integrated Pest Management IPM programs, we develop a generic model to study the impact of mating disruption control using an artificial female pheromone to confuse males and adversely affect their mating opportunities. Consequently the reproduction rate is diminished leading to a decline in the population size. For more efficient control, trapping is used to capture the males attracted to the artificial pheromone. The model, derived from biological and ecological assumptions, is governed by a system of ODEs. A theoretical analysis of the model without control is first carried out to establish the properties of the endemic equilibrium. Then, control is added and the theoretical analysis of the model enables to identify threshold values of pheromone which are practically interesting for field applications. In particular, we show that there is a threshold above which the global asymptotic stability of the trivial equilibrium is ensured, i.e. the population goes to extinction. Finally we illustrate the theoretical results via numerical experiments.

A String Model of Liquidity in Financial Markets ; We consider a dynamic market model of liquidity where unmatched buy and sell limit orders are stored in order books. The resulting net demand surface constitutes the sole input to the model. We prove that generically there is no arbitrage in the model when the driving noise is a stochastic string. Under the equivalent martingale measure, the clearing price is a martingale, and options can be priced under the noarbitrage hypothesis. We consider several parameterized versions of the model, and show some advantages of specifying the demand curve as quantity as a function of price as opposed to price as a function of quantity. We calibrate our model to real order book data, compute option prices by Monte Carlo simulation, and compare the results to observed data.

Modelling Chemical Reasoning to Predict Reactions ; The ability to reason beyond established knowledge allows Organic Chemists to solve synthetic problems and to invent novel transformations. Here, we propose a model which mimics chemical reasoning and formalises reaction prediction as finding missing links in a knowledge graph. We have constructed a knowledge graph containing 14.4 million molecules and 8.2 million binary reactions, which represents the bulk of all chemical reactions ever published in the scientific literature. Our model outperforms a rulebased expert system in the reaction prediction task for 180,000 randomly selected binary reactions. We show that our datadriven model generalises even beyond known reaction types, and is thus capable of effectively re discovering novel transformations even including transitionmetal catalysed reactions. Our model enables computers to infer hypotheses about reactivity and reactions by only considering the intrinsic local structure of the graph, and because each single reaction prediction is typically achieved in a subsecond time frame, our model can be used as a highthroughput generator of reaction hypotheses for reaction discovery.

Modeling epidemics on dcliqued graphs ; Since social interactions have been shown to lead to symmetric clusters, we propose here that symmetries play a key role in epidemic modeling. Mathematical models on dary tree graphs were recently shown to be particularly effective for modeling epidemics in simple networks Seibold Callender, 2016. To account for symmetric relations, we generalize this to a new type of networks modeled on dcliqued tree graphs, which are obtained by adding edges to regular dtrees to form dcliques. This setting gives a more realistic model for epidemic outbreaks originating, for example, within a family or classroom and which could reach a population by transmission via children in schools. Specifically, we quantify how an infection starting in a clique e.g. family can reach other cliques through the body of the graph e.g. public places. Moreover, we propose and study the notion of a safe zone, a subset that has a negligible probability of infection.

Randomized Block Cubic Newton Method ; We study the problem of minimizing the sum of three convex functions a differentiable, twicedifferentiable and a nonsmooth term in a high dimensional setting. To this effect we propose and analyze a randomized block cubic Newton RBCN method, which in each iteration builds a model of the objective function formed as the sum of the natural models of its three components a linear model with a quadratic regularizer for the differentiable term, a quadratic model with a cubic regularizer for the twice differentiable term, and perfect proximal model for the nonsmooth term. Our method in each iteration minimizes the model over a random subset of blocks of the search variable. RBCN is the first algorithm with these properties, generalizing several existing methods, matching the best known bounds in all special cases. We establish cal O1epsilon, cal O1sqrtepsilon and cal Olog 1epsilon rates under different assumptions on the component functions. Lastly, we show numerically that our method outperforms the stateoftheart on a variety of machine learning problems, including cubically regularized leastsquares, logistic regression with constraints, and Poisson regression.