Datasets:

sswt

/

arxiv_sample_50K

like 0

Learning Texture Invariant Representation for Domain Adaptation of Semantic Segmentation ; Since annotating pixellevel labels for semantic segmentation is laborious, leveraging synthetic data is an attractive solution. However, due to the domain gap between synthetic domain and real domain, it is challenging for a model trained with synthetic data to generalize to real data. In this paper, considering the fundamental difference between the two domains as the texture, we propose a method to adapt to the texture of the target domain. First, we diversity the texture of synthetic images using a style transfer algorithm. The various textures of generated images prevent a segmentation model from overfitting to one specific synthetic texture. Then, we finetune the model with selftraining to get direct supervision of the target texture. Our results achieve stateoftheart performance and we analyze the properties of the model trained on the stylized dataset with extensive experiments.

Optimal Regularization Can Mitigate Double Descent ; Recent empirical and theoretical studies have shown that many learning algorithms from linear regression to neural networks can have test performance that is nonmonotonic in quantities such the sample size and model size. This striking phenomenon, often referred to as double descent, has raised questions of if we need to rethink our current understanding of generalization. In this work, we study whether the doubledescent phenomenon can be avoided by using optimal regularization. Theoretically, we prove that for certain linear regression models with isotropic data distribution, optimallytuned ell2 regularization achieves monotonic test performance as we grow either the sample size or the model size. We also demonstrate empirically that optimallytuned ell2 regularization can mitigate double descent for more general models, including neural networks. Our results suggest that it may also be informative to study the test risk scalings of various algorithms in the context of appropriately tuned regularization.

An analytical approximation of the scalar spectrum in the ultraslowroll inflationary models ; The ultraslowroll USR inflationary models predict largeamplitude scalar perturbations at small scales which can lead to the primordial black hole production and scalarinduced gravitational waves. In general scalar perturbations in the USR models can only be obtained using numerical method because the usual slowroll approximation breaks. In this work, we propose an analytical approach to estimate the scalar spectrum which is consistent with the numerical result. We find that the USR inflationary models predict a peak with powerlaw slopes in the scalar spectrum and energy spectrum of gravitational waves, and we derive the expression of the spectral indexes in terms of the inflationary potential. In turn, the inflationary potential near the USR regime can be reconstructed from the negative spectral index of the gravitational wave energy spectrum.

Dynamics of a predatorprey model with generalized Holling type functional response and mutual interference ; Mutual interference and prey refuge are important drivers of predatorprey dynamics. The exponent or degree of mutual interference has been under much debate in theoretical ecology. In the present work, we investigate the interplay of the mutual interference exponent, on the behavior of a predatorprey model with a generalized Holling type functional response. We investigate stability properties of the system and derive conditions for the occurrence of saddlenode and Hopfbifurcations. A sufficient condition for extinction of the prey species has also been derived for the model. In addition, we investigate the effect of a prey refuge on the population dynamics of the model and derive conditions for the prey refuge that would yield persistence of populations. We provide additional verification our analytical results via numerical simulations. Our findings are in accordance with classical experimental results in ecology by Gauss G. F 1934, that show that extinction of predator and prey populations is possible in a finite time period but that bringing in refuge can effectively cause persistence.

Dropout Strikes Back Improved Uncertainty Estimation via Diversity Sampling ; Uncertainty estimation for machine learning models is of high importance in many scenarios such as constructing the confidence intervals for model predictions and detection of outofdistribution or adversarially generated points. In this work, we show that modifying the sampling distributions for dropout layers in neural networks improves the quality of uncertainty estimation. Our main idea consists of two main steps computing datadriven correlations between neurons and generating samples, which include maximally diverse neurons. In a series of experiments on simulated and realworld data, we demonstrate that the diversification via determinantal point processesbased sampling achieves stateoftheart results in uncertainty estimation for regression and classification tasks. An important feature of our approach is that it does not require any modification to the models or training procedures, allowing straightforward application to any deep learning model with dropout layers.

Localized Flood DetectionWith Minimal Labeled Social Media Data Using Transfer Learning ; Social media generates an enormous amount of data on a daily basis but it is very challenging to effectively utilize the data without annotating or labeling it according to the target application. We investigate the problem of localized flood detection using the social sensing model Twitter in order to provide an efficient, reliable and accurate flood text classification model with minimal labeled data. This study is important since it can immensely help in providing the floodrelated updates and notifications to the city officials for emergency decision making, rescue operations, and early warnings, etc. We propose to perform the text classification using the inductive transfer learning method i.e pretrained language model ULMFiT and finetune it in order to effectively classify the floodrelated feeds in any new location. Finally, we show that using very little new labeled data in the target domain we can successfully build an efficient and high performing model for flood detection and analysis with humangenerated facts and observations from Twitter.

Explainable Deep Classification Models for Domain Generalization ; Conventionally, AI models are thought to trade off explainability for lower accuracy. We develop a training strategy that not only leads to a more explainable AI system for object classification, but as a consequence, suffers no perceptible accuracy degradation. Explanations are defined as regions of visual evidence upon which a deep classification network makes a decision. This is represented in the form of a saliency map conveying how much each pixel contributed to the network's decision. Our training strategy enforces a periodic saliencybased feedback to encourage the model to focus on the image regions that directly correspond to the groundtruth object. We quantify explainability using an automated metric, and using human judgement. We propose explainability as a means for bridging the visualsemantic gap between different domains where model explanations are used as a means of disentagling domain specific information from otherwise relevant features. We demonstrate that this leads to improved generalization to new domains without hindering performance on the original domain.

Modeling Streaming Potential in Porous and Fractured Media, Description and Benefits of the Effective Excess Charge Density Approach ; Selfpotential signals can be generated by different sources and can be decomposed in various contributions. Streming potential is the contribution due to the water flux in the subsurface and is of particular interest in hydrogeophysics and reservoir characterization. Being able to estimate water fluxes in porous and fractured media using streaming potential data relies on our understanding of the electrokinetic coupling at the mineralsolution interface and our capacity to understand, model, and upscale this phenomenon. Two main approaches have been proposed to predict streaming potential generation in geological media. One of these approaches is based on determining the excess charge which is effectively dragged in the medium by water flow. In this chapter, we describe how to model the streaming potential by considering this effective excess charge density, how it can be defined, calculated and upscaled. We provide a short overview of the theoretical basis of this approach and we describe different applications to both water saturated and partially saturated soils and fractured media.

Plane symmetric model in fR,T gravity ; A plane symmetric BianchiI model is explored in fR,T gravity, where R is the Ricci scalar and T is the trace of energymomentum tensor. The solutions are obtained with the consideration of a specific Hubble parameter which yields a constant deceleration parameter. The various evolutionary phases are identified under the constraints obtained for physically viable cosmological scenarios. Although a single primary matter source is taken, due to the coupling between matter and fR,T gravity, an additional matter source appears, which mimics a perfect fluid or exotic matter. The solutions are also extended to the case of a scalar field model. The kinematical behavior of the model remains independent of fR,T gravity. The physical behavior of the effective matter also remains the same as in general relativity. It is found that fR,T gravity can be a good alternative to the hypothetical candidates of dark energy to describe the present accelerating expansion of the universe.

Classifying Pole of Amplitude Using Deep Neural Network ; Most of exotic resonances observed in the past decade appear as peak structure near some threshold. These nearthreshold phenomena can be interpreted as genuine resonant states or enhanced threshold cusps. Apparently, there is no straightforward way of distinguishing the two structures. In this work, we employ the strength of deep feedforward neural network in classifying objects with almost similar features. We construct a neural network model with scattering amplitude as input and nature of pole causing the enhancement as output. The training data is generated by an Smatrix satisfying the unitarity and analyticity requirements. Using the separable potential model, we generate a validation data set to measure the network's predictive power. We find that our trained neural network model gives high accuracy when the cutoff parameter of the validation data is within 400800mbox MeV. As a final test, we use the Nijmegen partial wave and potential models for nucleonnucleon scattering and show that the network gives the correct nature of pole.

MultiLabel Text Classification using Attentionbased Graph Neural Network ; In MultiLabel Text Classification MLTC, one sample can belong to more than one class. It is observed that most MLTC tasks, there are dependencies or correlations among labels. Existing methods tend to ignore the relationship among labels. In this paper, a graph attention networkbased model is proposed to capture the attentive dependency structure among the labels. The graph attention network uses a feature matrix and a correlation matrix to capture and explore the crucial dependencies between the labels and generate classifiers for the task. The generated classifiers are applied to sentence feature vectors obtained from the text feature extraction network BiLSTM to enable endtoend training. Attention allows the system to assign different weights to neighbor nodes per label, thus allowing it to learn the dependencies among labels implicitly. The results of the proposed model are validated on five realworld MLTC datasets. The proposed model achieves similar or better performance compared to the previous stateoftheart models.

Multipartite Purification, Multiboundary Wormholes and Islands in AdS3CFT2 ; The holographic duals of Entanglement of Purification through the Entanglement Wedge Cross Section has been a welldiscussed topic in the literature recently. More general entanglement measures involving multipartite information and their holographic duals have also been proposed. On the other hand, the recent intriguing program reproducing the Page Curve in Black hole entropy using the notion of islands has also been an obvious issue of attraction. A toy model involving Multiboundary wormholes in AdS3 was able to capture many interesting facts about such calculations. In such a toy model, the notion of islands was intuitively connected to quantum error correction. We try to bridge the ideas of the two programs especially in AdS3CFT2 and give a description of the islands in terms of multipartite entanglement of purification. This clarifies a few simplified assumptions made while describing the toy model and also enables us to understand the familiar information paradox within the framework of the same model.

Correspondence between temporal correlations in time series, inverse problems, and the Spherical Model ; In this paper we employ methods from Statistical Mechanics to model temporal correlations in time series. We put forward a methodology based on the Maximum Entropy principle to generate ensembles of time series constrained to preserve part of the temporal structure of an empirical time series of interest. We show that a constraint on the lagone autocorrelation can be fully handled analytically, and corresponds to the well known Spherical Model of a ferromagnet. We then extend such a model to include constraints on more complex temporal correlations by means of perturbation theory, showing that this leads to substantial improvements in capturing the lagone autocorrelation in the variance. We apply our approach on synthetic data, and illustrate how it can be used to formulate expectations on the future values of a data generating process.

Convex Parameter Recovery for Interacting Marked Processes ; We introduce a new general modeling approach for multivariate discrete event data with categorical interacting marks, which we refer to as marked Bernoulli processes. In the proposed model, the probability of an event of a specific category to occur in a location may be influenced by past events at this and other locations. We do not restrict interactions to be positive or decaying over time as it is commonly adopted, allowing us to capture an arbitrary shape of influence from historical events, locations, and events of different categories. In our modeling, prior knowledge is incorporated by allowing general convex constraints on model parameters. We develop two parameter estimation procedures utilizing the constrained Least Squares LS and Maximum Likelihood ML estimation, which are solved using variational inequalities with monotone operators. We discuss different applications of our approach and illustrate the performance of proposed recovery routines on synthetic examples and a realworld police dataset.

Analytical Model of MemoryBound Applications Compiled with High Level Synthesis ; The increasing demand of dedicated accelerators to improve energy efficiency and performance has highlighted FPGAs as a promising option to deliver both. However, programming FPGAs in hardware description languages requires long time and effort to achieve optimal results, which discourages many programmers from adopting this technology. High Level Synthesis tools improve the accessibility to FPGAs, but the optimization process is still time expensive due to the large compilation time, between minutes and days, required to generate a single bitstream. Whereas placing and routing take most of this time, the RTL pipeline and memory organization are known in seconds. This early information about the organization of the upcoming bitstream is enough to provide an accurate and fast performance model. This paper presents a performance analytical model for HLS designs focused on memory bound applications. With a careful analysis of the generated memory architecture and DRAM organization, the model predicts the execution time with a maximum error of 9.2 for a set of representative applications. Compared with previous works, our predictions reduce on average at least 2times the estimation error.

Inflationary Solution of Hamilton Jacobi Equations during Weak Dissipative Regime ; In this paper, an elegant mathematical approach is introduced to solve the equations of warm inflationary model without using extra approximations other than slowroll. This important inflationary method known as HamiltonJacobian formalism. Here tachyon field and the imperfect fluid are considered to be the cosmic ingredients to create inflation. A general formalism is developed for the considered inflationary model and further work is restricted to weak dissipative regime. A detailed analysis of the model is presented for three different choices of bulk and dissipative coefficients taking as constant as well as variable function of Hubble parameter and inflaton. In each case, the involved model parameters are constrained to plot the physical acceptable range of scalar spectral index and tensor to scalar ratio. The parametric trajectories proved that the acquired results for all the three cases are compatible with Planck astrophysical data. Furthermore, the existence of warm inflation and slowroll limit are also verified graphically.

PrivacyAware Data CleaningasaService Extended Version ; Data cleaning is a pervasive problem for organizations as they try to reap value from their data. Recent advances in networking and cloud computing technology have fueled a new computing paradigm called DatabaseasaService, where data management tasks are outsourced to large service providers. In this paper, we consider a Data CleaningasaService model that allows a client to interact with a data cleaning provider who hosts curated, and sensitive data. We present PACAS a PrivacyAware data CleaningAsaService model that facilitates interaction between the parties with client query requests for data, and a service provider using a data pricing scheme that computes prices according to data sensitivity. We propose new extensions to the model to define generalized data repairs that obfuscate sensitive data to allow data sharing between the client and service provider. We present a new semantic distance measure to quantify the utility of such repairs, and we redefine the notion of consistency in the presence of generalized values. The PACAS model uses X,Y,Lanonymity that extends existing data publishing techniques to consider the semantics in the data while protecting sensitive values. Our evaluation over real data show that PACAS safeguards semantically related sensitive values, and provides lower repair errors compared to existing privacyaware cleaning techniques.

A Simple Equivalent Circuit Approach for Anisotropic Frequency Selective Surfaces and Metasurfaces ; An equivalent circuit model for Frequency Selective Surfaces FSS comprising anisotropic elements is presented. The periodic surface is initially simulated with an arbitrary azimuthal incidence angle and its surface impedance matrix is derived. The impedance matrix is subsequently rotated by an angle varphirot on the crystal axes chi1, chi2 thus nullifying its extra diagonal terms. The rotation angle varphirot is derived according to the spectral theorem by using the terms of the matrix initially extracted. The diagonal terms of the rotated matrix, that is, the impedances Zchi1 and Zchi2, are finally matched with simple LC networks. The circuit model representation of the anisotropic element can be used to analyse anisotropic FSSs rotated by a generic azimuth angle. The methodology provides a compact description of generic FSS elements with only five parameters the lumped parameters of the LC network Lchi1, Cchi1, Lchi2, Cchi2 and the rotation angle varphirot. The circuit model can take into account the presence of dielectric substrates close to the FSS or a variation of the FSS periodicity without additional computational efforts. The equivalent circuit model is finally applied to the design of two transmitting polarization converts based on anisotropic metasurfaces.

A note on tools for prediction under uncertainty and identifiability of SIRlike dynamical systems for epidemiology ; We provide an overview of the methods that can be used for prediction under uncertainty and data fitting of dynamical systems, and of the fundamental challenges that arise in this context. The focus is on SIRlike models, that are being commonly used when attempting to predict the trend of the COVID19 pandemic. In particular, we raise a warning flag about identifiability of the parameters of SIRlike models; often, it might be hard to infer the correct values of the parameters from data, even for very simple models, making it nontrivial to use these models for meaningful predictions. Most of the points that we touch upon are actually generally valid for inverse problems in more general setups.

Automatic Composition of Guitar Tabs by Transformers and Groove Modeling ; Deep learning algorithms are increasingly developed for learning to compose music in the form of MIDI files. However, whether such algorithms work well for composing guitar tabs, which are quite different from MIDIs, remain relatively unexplored. To address this, we build a model for composing fingerstyle guitar tabs with TransformerXL, a neural sequence model architecture. With this model, we investigate the following research questions. First, whether the neural net generates note sequences with meaningful notestring combinations, which is important for the guitar but not other instruments such as the piano. Second, whether it generates compositions with coherent rhythmic groove, crucial for fingerstyle guitar music. And, finally, how pleasant the composed music is in comparison to real, humanmade compositions. Our work provides preliminary empirical evidence of the promise of deep learning for tab composition, and suggests areas for future study.

Exchangeable Neural ODE for Set Modeling ; Reasoning over an instance composed of a set of vectors, like a point cloud, requires that one accounts for intraset dependent features among elements. However, since such instances are unordered, the elements' features should remain unchanged when the input's order is permuted. This property, permutation equivariance, is a challenging constraint for most neural architectures. While recent work has proposed global pooling and attentionbased solutions, these may be limited in the way that intradependencies are captured in practice. In this work we propose a more general formulation to achieve permutation equivariance through ordinary differential equations ODE. Our proposed module, Exchangeable Neural ODE ExNODE, can be seamlessly applied for both discriminative and generative tasks. We also extend set modeling in the temporal dimension and propose a VAE based model for temporal set modeling. Extensive experiments demonstrate the efficacy of our method over strong baselines.

A Multilingual Neural Machine Translation Model for Biomedical Data ; We release a multilingual neural machine translation model, which can be used to translate text in the biomedical domain. The model can translate from 5 languages French, German, Italian, Korean and Spanish into English. It is trained with large amounts of generic and biomedical data, using domain tags. Our benchmarks show that it performs near stateoftheart both on news generic domain and biomedical test sets, and that it outperforms the existing publicly released models. We believe that this release will help the largescale multilingual analysis of the digital content of the COVID19 crisis and of its effects on society, economy, and healthcare policies. We also release a test set of biomedical text for KoreanEnglish. It consists of 758 sentences from official guidelines and recent papers, all about COVID19.

BoppPodolsky Scalar Electrodynamics Propagators and EnergyMomentum Tensor in Covariant and LightFront Coordinates ; We consider the interaction between a charged scalar boson and the BoppPodolsky gauge fields. The BoppPodolsky BP electrodynamics possesses both massive and massless propagation modes for the photon, while preserving gauge invariance. We obtain the propagator of all fields present in the model for the higherorder generalizations of the linear covariant, lightfront and doubly transverse lightfront gauges. Although BP's original model is described by a higherorder derivatives Lagrangian, it is possible to work with an equivalent reducedorder version by means of the introduction of an auxiliary vector field. We compute the gaugeinvariant improved energymomentum tensor for the full reducedorder interacting BP model. Besides the more traditional frontform view, we also discuss the lightfront perspective in both versions of the model. Within a Lagrangian framework approach we maintain explicit covariance at all steps and show that the field propagators, as well as the energymomentum tensor, can be cast into a lightfront closed form using specific properties of general coordinate transformations.

Universal Analytic Model of Irradiation Defect Dynamics in SilicaSilicon Structures ; Irradiation damage is a key physics issue for semiconductor devices under extreme environments. For decades, the ionizationirradiationinduced damage in transistors with silicasilicon structures under constant dose rate is modeled by a uniform generation of E' centers in the bulk silica region and their irreversible conversion to Pb centers at the silicasilicon interface. But, the traditional model fails to explain experimentally observed dependence of the defect concentrations on dose, especially at low dose rate. Here, we propose that, the generation of E' is decelerated due to the dispersive diffusion of induced holes in the disordered silica and the conversion of Pb is reversible due to recombinationenhanced defect reactions under irradiation. It is shown that the derived analytic model based on these new understandings can consistently explain the fundamental but puzzling dependence of the defect concentrations on dose and dose rate in a wide range.

Linguisticallyaware Attention for Reducing the SemanticGap in VisionLanguage Tasks ; Attention models are widely used in Visionlanguage VL tasks to perform the visualtextual correlation. Humans perform such a correlation with a strong linguistic understanding of the visual world. However, even the best performing attention model in VL tasks lacks such a highlevel linguistic understanding, thus creating a semantic gap between the modalities. In this paper, we propose an attention mechanism Linguisticallyaware Attention LAT that leverages object attributes obtained from generic object detectors along with pretrained language models to reduce this semantic gap. LAT represents visual and textual modalities in a common linguisticallyrich space, thus providing linguistic awareness to the attention process. We apply and demonstrate the effectiveness of LAT in three VL tasks CountingVQA, VQA, and Image captioning. In CountingVQA, we propose a novel countingspecific VQA model to predict an intuitive count and achieve stateoftheart results on five datasets. In VQA and Captioning, we show the generic nature and effectiveness of LAT by adapting it into various baselines and consistently improving their performance.

HpRNet Incorporating Residual Noise Modeling for Violin in a Variational Parametric Synthesizer ; Generative Models for Audio Synthesis have been gaining momentum in the last few years. More recently, parametric representations of the audio signal have been incorporated to facilitate better musical control of the synthesized output. In this work, we investigate a parametric model for violin tones, in particular the generative modeling of the residual bow noise to make for more natural tone quality. To aid in our analysis, we introduce a dataset of Carnatic Violin Recordings where bow noise is an integral part of the playing style of higher pitched notes in specific gestural contexts. We obtain insights about each of the harmonic and residual components of the signal, as well as their interdependence, via observations on the latent space derived in the course of variational encoding of the spectral envelopes of the sustained sounds.

Topological Defects on the Lattice Dualities and Degeneracies ; We construct topological defects in twodimensional classical lattice models and quantum chains. The defects satisfy local commutation relations guaranteeing that the partition function is independent of their path. These relations and their solutions are extended to allow defect lines to fuse, branch and satisfy all the properties of a fusion category. We show how the twodimensional classical lattice models and their topological defects are naturally described by boundary conditions of a TuraevViroBarrettWestbury partition function. These defects allow KramersWannier duality to be generalized to a large class of models, explaining exact degeneracies between nonsymmetryrelated ground states as well as in the lowenergy spectrum. They give a precise and general notion of twisted boundary conditions and the universal behaviour under Dehn twists. Gluing a topological defect to a boundary yields linear identities between partition functions with different boundary conditions, allowing ratios of the universal gfactor to be computed exactly on the lattice. We develop this construction in detail in a variety of examples, including the Potts, parafermion and height models.

A Value of Information Framework for Latent Variable Models ; In this paper, a general value of information VoI framework is formalised for latent variable models. In particular, the mutual information between the current status at the source node and the observed noisy measurements at the destination node is used to evaluate the information value, which gives the theoretical interpretation of the reduction in uncertainty in the current status given that we have measurements of the latent process. Moreover, the VoI expression for a hidden Markov model is obtained in this setting. Numerical results are provided to show the relationship between the VoI and the traditional age of information AoI metric, and the VoI of Markov and hidden Markov models are analysed for the particular case when the latent process is an OrnsteinUhlenbeck process. While the contributions of this work are theoretical, the proposed VoI framework is general and useful in designing wireless systems that support timely, but noisy, status updates in the physical world.

Varyingcoefficient stochastic differential equations with applications in ecology ; Stochastic differential equations SDEs are popular tools to analyse time series data in many areas, such as mathematical finance, physics, and biology. They provide a mechanistic description of the phenomeon of interest, and their parameters often have a clear interpretation. These advantages come at the cost of requiring a relatively simple model specification. We propose a flexible model for SDEs with timevarying dynamics where the parameters of the process are nonparametric functions of covariates, similar to generalized additive models. Combining the SDEs and nonparametric approaches allows for the SDE to capture more detailed, nonstationary, features of the datagenerating process. We present a computationally efficient method of approximate inference, where the SDE parameters can vary according to fixed covariate effects, random effects, or basispenalty smoothing splines. We demonstrate the versatility and utility of this approach with three applications in ecology, where there is often a modelling tradeoff between interpretability and flexibility.

GRIT Generative Rolefiller Transformers for Documentlevel Event Entity Extraction ; We revisit the classic problem of documentlevel rolefiller entity extraction REE for template filling. We argue that sentencelevel approaches are illsuited to the task and introduce a generative transformerbased encoderdecoder framework GRIT that is designed to model context at the document level it can make extraction decisions across sentence boundaries; is implicitly aware of noun phrase coreference structure, and has the capacity to respect crossrole dependencies in the template structure. We evaluate our approach on the MUC4 dataset, and show that our model performs substantially better than prior work. We also show that our modeling choices contribute to model performance, e.g., by implicitly capturing linguistic knowledge such as recognizing coreferent entity mentions.

Cosmological Constraints on Entropic Cosmology with Matter Creation ; We investigate entropic force cosmological models with the possibility of matter creation and energy exchange between the bulk and the horizon of a homogeneous and an isotropic flat Universe. We consider three different kinds of entropy, Bekenstein's, the nonextensive TsallisCirto's and the quartic entropy, plus some phenomenological functional forms for matter creation rate to model different entropic force models and put the observational constraints on them. We show that while most of them are basically indistinguishable from a standard LambdaCDM scenario, the Bekenstein entropic force model with a matter creation rate proportional to the Hubble parameter is statistically highly favored over LambdaCDM. As a general result, we also find that both the Hawking temperature parameter gamma, which relates the energy exchange between the bulk and the boundary of the Universe, and the matter creation rate Gammat, must be very small in order to reproduce observational data.

Transforming Probabilistic Programs for Model Checking ; Probabilistic programming is perfectly suited to reliable and transparent data science, as it allows the user to specify their models in a highlevel language without worrying about the complexities of how to fit the models. Static analysis of probabilistic programs presents even further opportunities for enabling a highlevel style of programming, by automating timeconsuming and errorprone tasks. We apply static analysis to probabilistic programs to automate large parts of two crucial model checking methods Prior Predictive Checks and SimulationBased Calibration. Our method transforms a probabilistic program specifying a density function into an efficient forwardsampling form. To achieve this transformation, we extract a factor graph from a probabilistic program using static analysis, generate a set of proposal directed acyclic graphs using a SAT solver, select a graph which will produce provably correct sampling code, then generate one or more sampling programs. We allow minimal user interaction to broaden the scope of application beyond what is possible with static analysis alone. We present an implementation targeting the popular Stan probabilistic programming language, automating large parts of a robust Bayesian workflow for a wide community of probabilistic programming users.

A generalized q growth model based on nonadditive entropy ; We present a general growth model based on nonextensive statistical physics is presented. The obtained equation is expressed in terms of nonadditive q entropy. We show that the most common unidimensional growth laws such as power law, exponential, logistic, Richards, Von Bertalanffy, Gompertz can be obtained. This model belongs as a particular case reported in Physica A 369, 645 2006. The new evolution equation resembles the universality revealed by West for ontogenetic growth Nature 413, 628 2001. We show that for early times the model follows a power law growth as Nt approx t D , where the exponent D equiv frac11q classifies different types of growth. Several examples are given and discussed.

Path Dependent Structural Equation Models ; Causal analyses of longitudinal data generally assume that the qualitative causal structure relating variables remains invariant over time. In structured systems that transition between qualitatively different states in discrete time steps, such an approach is deficient on two fronts. First, timevarying variables may have statespecific causal relationships that need to be captured. Second, an intervention can result in state transitions downstream of the intervention different from those actually observed in the data. In other words, interventions may counterfactually alter the subsequent temporal evolution of the system. We introduce a generalization of causal graphical models, Path Dependent Structural Equation Models PDSEMs, that can describe such systems. We show how causal inference may be performed in such models and illustrate its use in simulations and data obtained from a septoplasty surgical procedure.

Surrogate Model For Field Optimization Using BetaVAE Based Regression ; Oilfield development related decisions are made using reservoir simulationbased optimization study in which different production scenarios and well controls are compared. Such simulations are computationally expensive and so surrogate models are used to accelerate studies. Deep learning has been used in past to generate surrogates, but such models often fail to quantify prediction uncertainty and are not interpretable. In this work, betaVAE based regression is proposed to generate simulation surrogates for use in optimization workflow. betaVAE enables interpretable, factorized representation of decision variables in latent space, which is then further used for regression. Probabilistic dense layers are used to quantify prediction uncertainty and enable approximate Bayesian inference. Surrogate model developed using betaVAE based regression finds interpretable and relevant latent representation. A reasonable value of beta ensures a good balance between factor disentanglement and reconstruction. Probabilistic dense layer helps in quantifying predicted uncertainty for objective function, which is then used to decide whether fullphysics simulation is required for a case.

CompetenceBased Student Modelling with Dynamic Bayesian Networks ; We present a general method for using a competences map, created by defining generalizationspecialization and inclusionpartof relationships between competences, in order to build an overlay student model in the form of a dynamic Bayesian network in which conditional probability distributions are defined per relationship type. We have created a competences map for a subset of the transversal competences defined as educational goals for the Mexican high school system, then we have built a dynamic Bayesian student model as said before, and we have use it to trace the development of the corresponding competences by some hypothetical students exhibiting representative performances along an online course low to medium performance, medium to high performance but with low final score, and two terms medium to high performance. The results obtained suggest that the proposed way for constructing dynamic Bayesian student models on the basis of competences maps could be useful to monitor competence development by real students in online course.

31 Formulation of the StandardModel Extension Gravity Sector ; We present a 31 formulation of the effective field theory framework called the StandardModel Extension in the gravitational sector. The explicit local Lorentz and diffeomorphism symmetry breaking assumption is adopted and we perform a DiracHamiltonian analysis. We show that the structure of the dynamics presents significant differences from General Relativity and other modified gravity models. We explore Hamilton's equations for some special choices of the coefficients. Our main application is cosmology and we present the modified Friedmann equations for this case. The results show some intriguing modifications to standard cosmology. In addition, we compare our results to existing frameworks and models and we comment on the potential impact to other areas of gravitational theory and phenomenology.

Video Captioning Using Weak Annotation ; Video captioning has shown impressive progress in recent years. One key reason of the performance improvements made by existing methods lie in massive paired videosentence data, but collecting such strong annotation, i.e., highquality sentences, is timeconsuming and laborious. It is the fact that there now exist an amazing number of videos with weak annotation that only contains semantic concepts such as actions and objects. In this paper, we investigate using weak annotation instead of strong annotation to train a video captioning model. To this end, we propose a progressive visual reasoning method that progressively generates fine sentences from weak annotations by inferring more semantic concepts and their dependency relationships for video captioning. To model concept relationships, we use dependency trees that are spanned by exploiting external knowledge from large sentence corpora. Through traversing the dependency trees, the sentences are generated to train the captioning model. Accordingly, we develop an iterative refinement algorithm that refines sentences via spanning dependency trees and finetunes the captioning model using the refined sentences in an alternative training manner. Experimental results demonstrate that our method using weak annotation is very competitive to the stateoftheart methods using strong annotation.

A fully basis invariant Symmetry Map of the 2HDM ; We derive necessary and sufficient conditions for all global symmetries of the most general two Higgs doublet model 2HDM scalar potential entirely in terms of reparametrization independent, i.e. basis invariant, objects. This culminates in what we call a Symmetry Map of the parameter space of the model and the fundamental insight that there are, in general, two algebraically distinct ways of how symmetries manifest themselves on basis invariant objects either, basis invariant objects can be nontrivially related, or, basis covariant objects can vanish. These two options have different consequences on the resulting structure of the ring of basis invariants and on the number of remaining physical parameters. Alongside, we derive for the first time necessary and sufficient conditions for CP conservation in the 2HDM entirely in terms of CPeven quantities. This study lays the methodological foundation for analogous investigations of global symmetries in all other models that have unphysical freedom of reparametrization, most notably the Standard Model flavor sector.

Inverse square root levelcrossing quantum twostate model ; We introduce a new unconditionally solvable levelcrossing twostate model given by a constantamplitude optical field configuration for which the detuning is an inversesquareroot function of time. This is a member of one of the five families of biconfluent Heun models. We prove that this is the only nonclassical exactly solvable field configuration among the biconfluent Heun classes, solvable in terms of finite sums of the Hermite functions. The general solution of the twostate problem for this model is written in terms of four Hermite functions of a shifted and scaled argument each of the two fundamental solutions presents an irreducible combination of two Hermite functions. We present the general solution, rewrite it in terms of more familiar physical quantities and analyze the time dynamics of a quantum system subject to excitation by a laser field of this configuration.

The WellTempered Cosmological Constant The Horndeski Variations ; Well tempering is one of the few classical field theory methods for solving the original cosmological constant problem, dynamically canceling a large possibly Planck scale vacuum energy and leaving the matter component intact, while providing a viable cosmology with late time cosmic acceleration and an end de Sitter state. We present the general constraints that variations of Horndeski gravity models with different combinations of terms must satisfy to admit an exact de Sitter spacetime that does not respond to an arbitrarily large cosmological constant. We explicitly derive several specific scalartensor models that well temper and can deliver a standard cosmic history including current cosmic acceleration. Stability criteria, attractor behavior of the de Sitter state, and the response of the models to pressureless matter are considered. The well tempered conditions can be used to focus on particular models of modified gravity that have special interest not only removing the original cosmological constant problem but providing relations between the free Horndeski functions and reducing them to a couple of parameters, suitable for testing gravity and cosmological data analysis.

Adversarial Watermarking Transformer Towards Tracing Text Provenance with Data Hiding ; Recent advances in natural language generation have introduced powerful language models with highquality output text. However, this raises concerns about the potential misuse of such models for malicious purposes. In this paper, we study natural language watermarking as a defense to help better mark and trace the provenance of text. We introduce the Adversarial Watermarking Transformer AWT with a jointly trained encoderdecoder and adversarial training that, given an input text and a binary message, generates an output text that is unobtrusively encoded with the given message. We further study different training and inference strategies to achieve minimal changes to the semantics and correctness of the input text. AWT is the first endtoend model to hide data in text by automatically learning without ground truth word substitutions along with their locations in order to encode the message. We empirically show that our model is effective in largely preserving text utility and decoding the watermark while hiding its presence against adversaries. Additionally, we demonstrate that our method is robust against a range of attacks.

Learning more expressive joint distributions in multimodal variational methods ; Data often are formed of multiple modalities, which jointly describe the observed phenomena. Modeling the joint distribution of multimodal data requires larger expressive power to capture highlevel concepts and provide better data representations. However, multimodal generative models based on variational inference are limited due to the lack of flexibility of the approximate posterior, which is obtained by searching within a known parametric family of distributions. We introduce a method that improves the representational capacity of multimodal variational methods using normalizing flows. It approximates the joint posterior with a simple parametric distribution and subsequently transforms into a more complex one. Through several experiments, we demonstrate that the model improves on stateoftheart multimodal methods based on variational inference on various computer vision tasks such as colorization, edge and mask detection, and weakly supervised learning. We also show that learning more powerful approximate joint distributions improves the quality of the generated samples. The code of our model is publicly available at httpsgithub.comSashoNedelkoskiBPFDMVM.

Kernelbased parameter estimation of dynamical systems with unknown observation functions ; A lowdimensional dynamical system is observed in an experiment as a highdimensional signal; for example, a video of a chaotic pendulums system. Assuming that we know the dynamical model up to some unknown parameters, can we estimate the underlying system's parameters by measuring its timeevolution only once The key information for performing this estimation lies in the temporal interdependencies between the signal and the model. We propose a kernelbased score to compare these dependencies. Our score generalizes a maximum likelihood estimator for a linear model to a general nonlinear setting in an unknown feature space. We estimate the system's underlying parameters by maximizing the proposed score. We demonstrate the accuracy and efficiency of the method using two chaotic dynamical systems the double pendulum and the Lorenz '63 model.

TextIndependent Speaker Verification with Dual Attention Network ; This paper presents a novel design of attention model for textindependent speaker verification. The model takes a pair of input utterances and generates an utterancelevel embedding to represent speakerspecific characteristics in each utterance. The input utterances are expected to have highly similar embeddings if they are from the same speaker. The proposed attention model consists of a selfattention module and a mutual attention module, which jointly contributes to the generation of the utterancelevel embedding. The selfattention weights are computed from the utterance itself while the mutualattention weights are computed with the involvement of the other utterance in the input pairs. As a result, each utterance is represented by a selfattention weighted embedding and a mutualattention weighted embedding. The similarity between the embeddings is measured by a cosine distance score and a binary classifier output score. The whole model, named Dual Attention Network, is trained endtoend on Voxceleb database. The evaluation results on Voxceleb 1 test set show that the Dual Attention Network significantly outperforms the baseline systems. The best result yields an equal error rate of 16.

On Robustness and Bias Analysis of BERTbased Relation Extraction ; Finetuning pretrained models have achieved impressive performance on standard natural language processing benchmarks. However, the resultant model generalizability remains poorly understood. We do not know, for example, how excellent performance can lead to the perfection of generalization models. In this study, we analyze a finetuned BERT model from different perspectives using relation extraction. We also characterize the differences in generalization techniques according to our proposed improvements. From empirical experimentation, we find that BERT suffers a bottleneck in terms of robustness by way of randomizations, adversarial and counterfactual tests, and biases i.e., selection and semantic. These findings highlight opportunities for future improvements. Our opensourced testbed DiagnoseRE is available in urlhttpsgithub.comzjunlpDiagnoseRE.

MSRDARTS Minimum Stable Rank of Differentiable Architecture Search ; In neural architecture search NAS, differentiable architecture search DARTS has recently attracted much attention due to its high efficiency. It defines an overparameterized network with mixed edges, each of which represents all operator candidates, and jointly optimizes the weights of the network and its architecture in an alternating manner. However, this method finds a model with the weights converging faster than the others, and such a model with fastest convergence often leads to overfitting. Accordingly, the resulting model cannot always be wellgeneralized. To overcome this problem, we propose a method called minimum stable rank DARTS MSRDARTS, for finding a model with the best generalization error by replacing architecture optimization with the selection process using the minimum stable rank criterion. Specifically, a convolution operator is represented by a matrix, and MSRDARTS selects the one with the smallest stable rank. We evaluated MSRDARTS on CIFAR10 and ImageNet datasets. It achieves an error rate of 2.54 with 4.0M parameters within 0.3 GPUdays on CIFAR10, and a top1 error rate of 23.9 on ImageNet. The official code is available at httpsgithub.commtaecchhimsrdarts.git.

The mass splitting in an 331TC coupled Scenario ; The root of most of the technicolor TC problems lies in the way the ordinary fermions acquire their masses, where an ordinary fermion f couples to a technifermion F mediated by an Extended Technicolor ETC boson leading to fermion masses that vary with the ETC mass scale ME as 1ME2. Recently, we discussed a new approach consisting of models where TC and QCD are coupled through a larger theory, in this case the solutions of these equations are modified compared to those of the isolated equations, and TC and QCD selfenergies are of the Irregular form, which allows us to build models where ETC boson masses can be pushed to very high energies. In this work we extend these results for 331TC models, in particular considering a coupled system of SchwingerDyson equations, we show that all technifermions of the model exhibit the same asymptotic behavior for TC selfenergies. As an application we discuss how the mass splitting of the order O100GeV could be generated between the second and third generation of fermions.

Extremal Indices in the Series Scheme and their Applications ; We generalize the concept of extremal index of a stationary random sequence to the series scheme of identically distributed random variables with random series sizes tending to infinity in probability. We introduce new extremal indices through two definitions generalizing the basic properties of the classical extremal index. We prove some useful properties of the new extremal indices. We show how the behavior of aggregate activity maxima on random graphs in information network models and the behavior of maxima of random particle scores in branching processes in biological population models can be described in terms of the new extremal indices. We also obtain new results on models with copulas and threshold models. We show that the new indices can take different values for the same system, as well as values greater than one.

PrivacyPreserving Machine Learning Training in Aggregation Scenarios ; To develop Smart City, the growing popularity of Machine Learning ML that appreciates highquality training datasets generated from diverse IoT devices raises natural questions about the privacy guarantees that can be provided in such settings. Privacypreserving ML training in an aggregation scenario enables a model demander to securely train ML models with the sensitive IoT data gathered from personal IoT devices. Existing solutions are generally serveraided, cannot deal with the collusion threat between the servers or between the servers and data owners, and do not match the delicate environments of IoT. We propose a privacypreserving ML training framework named Heda that consists of a library of building blocks based on partial homomorphic encryption PHE enabling constructing multiple privacypreserving ML training protocols for the aggregation scenario without the assistance of untrusted servers and defending the security under collusion situations. Rigorous security analysis demonstrates the proposed protocols can protect the privacy of each participant in the honestbutcurious model and defend the security under most collusion situations. Extensive experiments validate the efficiency of Heda which achieves the privacypreserving ML training without losing the model accuracy.

Data Augmentation for Graph Classification ; Graph classification, which aims to identify the category labels of graphs, plays a significant role in drug classification, toxicity detection, protein analysis etc. However, the limitation of scale of benchmark datasets makes it easy for graph classification models to fall into overfitting and undergeneralization. Towards this, we introduce data augmentation on graphs and present two heuristic algorithms random mapping and motifsimilarity mapping, to generate more weakly labeled data for smallscale benchmark datasets via heuristic modification of graph structures. Furthermore, we propose a generic model evolution framework, MEvolve, which combines graph augmentation, data filtration and model retraining to optimize pretrained graph classifiers. Experiments conducted on six benchmark datasets demonstrate that MEvolve helps existing graph classification models alleviate overfitting when training on smallscale benchmark datasets and yields an average improvement of 312 accuracy on graph classification tasks.

Selection of Regression Models under Linear Restrictions for Fixed and Random Designs ; Many important modeling tasks in linear regression, including variable selection in which slopes of some predictors are set equal to zero and simplified models based on sums or differences of predictors in which slopes of those predictors are set equal to each other, or the negative of each other, respectively, can be viewed as being based on imposing linear restrictions on regression parameters. In this paper, we discuss how such models can be compared using information criteria designed to estimate predictive measures like squared error and KullbackLeibler KL discrepancy, in the presence of either deterministic predictors fixedX or random predictors randomX. We extend the justifications for existing fixedX criteria Cp, FPE and AICc, and randomX criteria Sp and RCp, to general linear restrictions. We further propose and justify a KLbased criterion, RAICc, under randomX for variable selection and general linear restrictions. We show in simulations that the use of the KLbased criteria AICc and RAICc results in better predictive performance and sparser solutions than the use of squared errorbased criteria, including crossvalidation.

Embeddingbased Zeroshot Retrieval through Query Generation ; Passage retrieval addresses the problem of locating relevant passages, usually from a large corpus, given a query. In practice, lexical termmatching algorithms like BM25 are popular choices for retrieval owing to their efficiency. However, termbased matching algorithms often miss relevant passages that have no lexical overlap with the query and cannot be finetuned to downstream datasets. In this work, we consider the embeddingbased twotower architecture as our neural retrieval model. Since labeled data can be scarce and because neural retrieval models require vast amounts of data to train, we propose a novel method for generating synthetic training data for retrieval. Our system produces remarkable results, significantly outperforming BM25 on 5 out of 6 datasets tested, by an average of 2.45 points for Recall1. In some cases, our model trained on synthetic data can even outperform the same model trained on real data

Complex Vehicle Routing with Memory Augmented Neural Networks ; Complex reallife routing challenges can be modeled as variations of wellknown combinatorial optimization problems. These routing problems have long been studied and are difficult to solve at scale. The particular setting may also make exact formulation difficult. Deep Learning offers an increasingly attractive alternative to traditional solutions, which mainly revolve around the use of various heuristics. Deep Learning may provide solutions which are less timeconsuming and of higher quality at large scales, as it generally does not need to generate solutions in an iterative manner, and Deep Learning models have shown a surprising capacity for solving complex tasks in recent years. Here we consider a particular variation of the Capacitated Vehicle Routing CVRP problem and investigate the use of Deep Learning models with explicit memory components. Such memory components may help in gaining insight into the model's decisions as the memory and operations on it can be directly inspected at any time, and may assist in scaling the method to such a size that it becomes viable for industry settings.

A ModelDriven Architecture Approach for Developing Healthcare ERP Case study in Morocco ; Nowadays, there are many problems in the Enterprise Resource Planning ERP implemented in the majority of hospitals in Morocco such as the difficulty of adaptation by the different users, the lack of several functionalities, errors that block the daily work, etc. All these problems require frequent modifications in the code, which implies a high effort to develop healthcare ERP as one of complex systems. In this paper, we are going to present a modeldriven approach for developing healthcare ERP based on class diagram. First, we constitute the independent model using UML, define the transformation rules then apply them on our source model class to generate at the end an XML file that will be necessary for the ERP code. Our approach will not only resolve the above problems, but also improve the efficiency of software development through the automatically generated code.

FocusConstrained Attention Mechanism for CVAEbased Response Generation ; To model diverse responses for a given post, one promising way is to introduce a latent variable into Seq2Seq models. The latent variable is supposed to capture the discourselevel information and encourage the informativeness of target responses. However, such discourselevel information is often too coarse for the decoder to be utilized. To tackle it, our idea is to transform the coarsegrained discourselevel information into finegrained wordlevel information. Specifically, we firstly measure the semantic concentration of corresponding target response on the post words by introducing a finegrained focus signal. Then, we propose a focusconstrained attention mechanism to take full advantage of focus in well aligning the input to the target response. The experimental results demonstrate that by exploiting the finegrained signal, our model can generate more diverse and informative responses compared with several stateoftheart models.

Robust Trajectory Optimization over Uncertain Terrain with Stochastic Complementarity ; Trajectory optimization with contactrich behaviors has recently gained attention for generating diverse locomotion behaviors without prespecified ground contact sequences. However, these approaches rely on precise models of robot dynamics and the terrain and are susceptible to uncertainty. Recent works have attempted to handle uncertainties in the system model, but few have investigated uncertainty in contact dynamics. In this study, we model uncertainty stemming from the terrain and design corresponding risksensitive objectives under the framework of contactimplicit trajectory optimization. In particular, we parameterize uncertainties from the terrain contact distance and friction coefficients using probability distributions and propose a corresponding expected residual minimization cost design approach. We evaluate our method in three simple robotic examples, including a legged hopping robot, and we benchmark one of our examples in simulation against a robust worstcase solution. We show that our risksensitive method produces contactaverse trajectories that are robust to terrain perturbations. Moreover, we demonstrate that the resulting trajectories converge to those generated by a traditional, nonrobust method as the terrain model becomes more certain. Our study marks an important step towards a fully robust, contactimplicit approach suitable for deploying robots on realworld terrain.

Unsupervised Model Adaptation for Continual Semantic Segmentation ; We develop an algorithm for adapting a semantic segmentation model that is trained using a labeled source domain to generalize well in an unlabeled target domain. A similar problem has been studied extensively in the unsupervised domain adaptation UDA literature, but existing UDA algorithms require access to both the source domain labeled data and the target domain unlabeled data for training a domain agnostic semantic segmentation model. Relaxing this constraint enables a user to adapt pretrained models to generalize in a target domain, without requiring access to source data. To this end, we learn a prototypical distribution for the source domain in an intermediate embedding space. This distribution encodes the abstract knowledge that is learned from the source domain. We then use this distribution for aligning the target domain distribution with the source domain distribution in the embedding space. We provide theoretical analysis and explain conditions under which our algorithm is effective. Experiments on benchmark adaptation task demonstrate our method achieves competitive performance even compared with joint UDA approaches.

Small Data, Big Decisions Model Selection in the SmallData Regime ; Highly overparametrized neural networks can display curiously strong generalization performance a phenomenon that has recently garnered a wealth of theoretical and empirical research in order to better understand it. In contrast to most previous work, which typically considers the performance as a function of the model size, in this paper we empirically study the generalization performance as the size of the training set varies over multiple orders of magnitude. These systematic experiments lead to some interesting and potentially very useful observations; perhaps most notably that training on smaller subsets of the data can lead to more reliable model selection decisions whilst simultaneously enjoying smaller computational costs. Our experiments furthermore allow us to estimate Minimum Description Lengths for common datasets given modern neural network architectures, thereby paving the way for principled model selection taking into account Occamsrazor.

LargeScale Parameters of SpatioTemporal ShortRange Indoor Backhaul Channels at 140 GHz ; The use of above100 GHz radio frequencies would be one of promising approaches to enhance the fifthgeneration cellular further. Any air interface and cellular network designs require channel models, for which measured evidence of largescale parameters such as pathloss, delay and angular spreads, is crucial. This paper provides the evidence from quasistatic spatiotemporal channel sounding campaigns at two indoor hotspot InH scenarios at 140 GHz band, assuming shortrange backhaul connectivity. The measured two InH sites are shopping mall and airport checkin hall. Our estimated omnidirectional largescale parameters from the measurements are found in good match with those of the Third Generation Partnership Project 3GPP for new radios NR channel model in InH scenario, despite the difference of assumed link types and radio frequency range. The 3GPP NR channel model is meant for access links and said to be valid up to 100 GHz, while our measurements cover shortrange backhaul scenarios at 140 GHz. We found more deviation between our estimated largescale parameters and those of the 3GPP NR channel model in the airport than in the shopping mall.

Observational constraints in Delta Gravity CMB and supernovas ; Delta Gravity is a gravitational model based on an extension of General Relativity given by a new symmetry called tildedelta. In this model, new matter fields are added to the original matter fields, motivated by the additional symmetry. We call them tildedelta matter fields. This model predicts an accelerating Universe without the need to introduce a cosmological constant. In this work, we study the Delta Gravity prediction about the scalar CMB TT power spectrum using an analytical hydrodynamical approach. To fit the Planck satellite's data with the DG model, we used a Markov Chain Monte Carlo analysis. We also include a study about the compatibility between SNeIa and CMB observations in the Delta Gravity Context. Finally, we obtain the scalar CMB TT power spectrum and the fitted parameters needed to explain both SNeIa Data and CMB measurements. The results are in a reasonable agreement with both observations considering the analytical approximation. We also discuss if the Hubble Constant and the Accelerating Universe are in concordance with the observational evidence in the Delta Gravity context.

Replica Analysis of the Linear Model with Markov or Hidden Markov Signal Priors ; This paper estimates free energy, average mutual information, and minimum mean square error MMSE of a linear model under two assumptions 1 the source is generated by a Markov chain, 2 the source is generated via a hidden Markov model. Our estimates are based on the replica method in statistical physics. We show that under the posterior mean estimator, the linear model with Markov sources or hidden Markov sources is decoupled into singleinput AWGN channels with state information available at both encoder and decoder where the state distribution follows the left PerronFrobenius eigenvector with unit Manhattan norm of the stochastic matrix of Markov chains. Numerical results show that the free energies and MSEs obtained via the replica method are closely approximate to their counterparts achieved by the MetropolisHastings algorithm or some wellknown approximate message passing algorithms in the research literature.

Resilience Analysis and Cascading FailureModeling of Power Systems under Extreme Temperatures ; In this paper, we propose an AC power flow based cascading failure model that explicitly considers external weather conditions, extreme temperatures in particular, and evaluates the impact of extreme temperature on the initiation and propagation of cascading blackouts. Specifically, load and dynamic line rating changes are modeled due to temperature disturbance, the probabilities for transmission line and generator outages are evaluated, and the timing for each type of events is carefully calculated to decide the actual event sequence. It should be emphasized that the correlated events, in the advent of external temperature changes, could together contribute to voltage instability. Besides, we model undervoltage load shedding and operator redispatch as control strategies for preventing the propagation of cascading failures. The effectiveness of the proposed model is verified by simulation results on the RTS96 3area system and it is found that temperature disturbances can lead to correlated load change and linegenerator tripping, which together will greatly increase the risk of cascading and voltage instability. Critical temperature change, critical area with temperature disturbance, identification of most vulnerable buses, and comparison of different control strategies are also carefully investigated.

On the periodicity of cardiovascular fluid dynamics simulations ; Threedimensional cardiovascular fluid dynamics simulations typically require computation of several cardiac cycles before they reach a periodic solution, rendering them computationally expensive. Furthermore, there is currently no standardized method to determine whether a simulation has yet reached that periodic state. In this work, we propose use of the asymptotic error measure to quantify the difference between simulation results and their ideal periodic state using lumpedparameter modeling. We further show that initial conditions are crucial in reducing computational time and develop an automated framework to generate appropriate initial conditions from a onedimensional model of blood flow. We demonstrate the performance of our initialization method using six patientspecific models from the Vascular Model Repository. In our examples, our initialization protocol achieves periodic convergence within one or two cardiac cycles, leading to a significant reduction in computational cost compared to standard methods. All computational tools used in this work are implemented in the opensource software platform SimVascular. Automatically generated initial conditions have the potential to significantly reduce computation time in cardiovascular fluid dynamics simulations.

A Stochastic Time Series Model for Predicting Financial Trends using NLP ; Stock price forecasting is a highly complex and vitally important field of research. Recent advancements in deep neural network technology allow researchers to develop highly accurate models to predict financial trends. We propose a novel deep learning model called STGAN, or Stochastic Timeseries Generative Adversarial Network, that analyzes both financial news texts and financial numerical data to predict stock trends. We utilize cuttingedge technology like the Generative Adversarial Network GAN to learn the correlations among textual and numerical data over time. We develop a new method of training a timeseries GAN directly using the learned representations of Naive Bayes' sentiment analysis on financial text data alongside technical indicators from numerical data. Our experimental results show significant improvement over various existing models and prior research on deep neural networks for stock price forecasting.

Accurate Prediction Of Machining Feedrate And Cycle Time Prediction Considering Interpolator Dynamics ; This paper presents an accurate machining feedrate prediction technique by modeling the trajectory generation behaviour of modern CNC machine tools. Typically, CAM systems simulate machines motion based on the commanded feedrate and the path geometry. Such approach does not consider the feed planning and interpolation strategy of the machines numerical control NC system. In this study, trajectory generation behaviour of the NC system is modelled and accurate cycle time prediction for complex machining toolpaths is realized. NC systems linear interpolation dynamics and commanded axis kinematic profiles are predicted by using Finite Impulse Response FIR based lowpass filters. The corner blending behaviour during nonstop interpolation of linear segments is modeled, and for the first time, the minimum cornering feedrate, that satisfies both the tolerance and machining constraints, has been calculated analytically for 3 axis toolpaths of any geometry. The proposed method is applied to 4 different case studies including complex machining toolpaths. Experimental validations show actual cycle times can be estimated with greater than 90 percent accuracy, greatly outperforming CAMbased predictions. It is expected that the proposed approach will help improve the accuracy of virtual machining models and support businesses decision making when costing machining processes.

Decoupling of Asymmetric Dark Matter During an Early Matter Dominated Era ; In models of Asymmetric Dark Matter ADM the relic density is set by a particle asymmetry in an analogous manner to the baryons. Here we explore the scenario in which ADM decouples from the Standard Model thermal bath during an early period of matter domination. We first present a model independent analysis for a generic ADM candidate with swave annihilation cross section with fairly general assumptions regarding the origin of the early matter dominated period. We contrast our results to those from conventional ADM models which assume radiation domination during decoupling. Subsequently, we examine an explicit example of this scenario in the context of an elegant SO10 implementation of ADM in which the matter dominated era is due to a long lived heavy righthanded neutrino. In the concluding remarks we discuss the prospects for superheavy ADM in this setting.

Adversarial Training Makes Weight Loss Landscape Sharper in Logistic Regression ; Adversarial training is actively studied for learning robust models against adversarial examples. A recent study finds that adversarially trained models degenerate generalization performance on adversarial examples when their weight loss landscape, which is loss changes with respect to weights, is sharp. Unfortunately, it has been experimentally shown that adversarial training sharpens the weight loss landscape, but this phenomenon has not been theoretically clarified. Therefore, we theoretically analyze this phenomenon in this paper. As a first step, this paper proves that adversarial training with the L2 norm constraints sharpens the weight loss landscape in the linear logistic regression model. Our analysis reveals that the sharpness of the weight loss landscape is caused by the noise aligned in the direction of increasing the loss, which is used in adversarial training. We theoretically and experimentally confirm that the weight loss landscape becomes sharper as the magnitude of the noise of adversarial training increases in the linear logistic regression model. Moreover, we experimentally confirm the same phenomena in ResNet18 with softmax as a more general case.

Adversarial example generation with AdaBelief Optimizer and Crop Invariance ; Deep neural networks are vulnerable to adversarial examples, which are crafted by applying small, humanimperceptible perturbations on the original images, so as to mislead deep neural networks to output inaccurate predictions. Adversarial attacks can thus be an important method to evaluate and select robust models in safetycritical applications. However, under the challenging blackbox setting, most existing adversarial attacks often achieve relatively low success rates on adversarially trained networks and advanced defense models. In this paper, we propose AdaBelief Iterative Fast Gradient Method ABIFGM and CropInvariant attack Method CIM to improves the transferability of adversarial examples. ABIFGM and CIM can be readily integrated to build a strong gradientbased attack to further boost the success rates of adversarial examples for blackbox attacks. Moreover, our method can also be naturally combined with other gradientbased attack methods to build a more robust attack to generate more transferable adversarial examples against the defense models. Extensive experiments on the ImageNet dataset demonstrate the method's effectiveness. Whether on adversarially trained networks or advanced defense models, our method has higher success rates than stateoftheart gradientbased attack methods.

A generalized model of flocking with steering ; We introduce and analyze a model for the dynamics of flocking and steering of a finite number of agents. In this model, each agent's acceleration consists of flocking and steering components. The flocking component is a generalization of many of the existing models and allows for the incorporation of many real world features such as acceleration bounds, partial masking effects and orientation bias. The steering component is also integral to capture real world phenomena. We provide rigorous sufficient conditions under which the agents flock and steer together. We also provide a formal singular perturbation study of the situation where flocking happens much faster than steering. We end our work by providing some numerical simulations to illustrate our theoretical results.

Argmax Flows and Multinomial Diffusion Learning Categorical Distributions ; Generative flows and diffusion models have been predominantly trained on ordinal data, for example natural images. This paper introduces two extensions of flows and diffusion for categorical data such as language or image segmentation Argmax Flows and Multinomial Diffusion. Argmax Flows are defined by a composition of a continuous distribution such as a normalizing flow, and an argmax function. To optimize this model, we learn a probabilistic inverse for the argmax that lifts the categorical data to a continuous space. Multinomial Diffusion gradually adds categorical noise in a diffusion process, for which the generative denoising process is learned. We demonstrate that our method outperforms existing dequantization approaches on text modelling and modelling on image segmentation maps in loglikelihood.

A Compositional Atlas of Tractable Circuit Operations From Simple Transformations to Complex InformationTheoretic Queries ; Circuit representations are becoming the lingua franca to express and reason about tractable generative and discriminative models. In this paper, we show how complex inference scenarios for these models that commonly arise in machine learning from computing the expectations of decision tree ensembles to informationtheoretic divergences of deep mixture models can be represented in terms of tractable modular operations over circuits. Specifically, we characterize the tractability of a vocabulary of simple transformations sums, products, quotients, powers, logarithms, and exponentials in terms of sufficient structural constraints of the circuits they operate on, and present novel hardness results for the cases in which these properties are not satisfied. Building on these operations, we derive a unified framework for reasoning about tractable models that generalizes several results in the literature and opens up novel tractable inference scenarios.

Distribution Free Uncertainty for the Minimum Norm Solution of Overparameterized Linear Regression ; A fundamental principle of learning theory is that there is a tradeoff between the complexity of a prediction rule and its ability to generalize. Modern machine learning models do not obey this paradigm They produce an accurate prediction even with a perfect fit to the training set. We investigate overparameterized linear regression models focusing on the minimum norm solution This is the solution with the minimal norm that attains a perfect fit to the training set. We utilize the recently proposed predictive normalized maximum likelihood pNML learner which is the minmax regret solution for the distributionfree setting. We derive an upper bound of this minmax regret which is associated with the prediction uncertainty. We show that if the test sample lies mostly in a subspace spanned by the eigenvectors associated with the large eigenvalues of the empirical correlation matrix of the training data, the model generalizes despite its overparameterized nature. We demonstrate the use of the pNML regret as a pointwise learnability measure on synthetic data and successfully observe the doubledecent phenomenon of the overparameterized models on UCI datasets.

Geostatistical Learning Challenges and Opportunities ; Statistical learning theory provides the foundation to applied machine learning, and its various successful applications in computer vision, natural language processing and other scientific domains. The theory, however, does not take into account the unique challenges of performing statistical learning in geospatial settings. For instance, it is well known that model errors cannot be assumed to be independent and identically distributed in geospatial a.k.a. regionalized variables due to spatial correlation; and trends caused by geophysical processes lead to covariate shifts between the domain where the model was trained and the domain where it will be applied, which in turn harm the use of classical learning methodologies that rely on random samples of the data. In this work, we introduce the geostatistical transfer learning problem, and illustrate the challenges of learning from geospatial data by assessing widelyused methods for estimating generalization error of learning models, under covariate shift and spatial correlation. Experiments with synthetic Gaussian process data as well as with real data from geophysical surveys in New Zealand indicate that none of the methods are adequate for model selection in a geospatial context. We provide general guidelines regarding the choice of these methods in practice while new methods are being actively researched.

TransMask A Compact and Fast Speech Separation Model Based on Transformer ; Speech separation is an important problem in speech processing, which targets to separate and generate clean speech from a mixed audio containing speech from different speakers. Empowered by the deep learning technologies over sequencetosequence domain, recent neural speech separation models are now capable of generating highly clean speech audios. To make these models more practical by reducing the model size and inference time while maintaining high separation quality, we propose a new transformerbased speech separation approach, called TransMask. By fully unleashing the power of selfattention on longterm dependency exception, we demonstrate the size of TransMask is more than 60 smaller and the inference is more than 2 times faster than stateoftheart solutions. TransMask fully utilizes the parallelism during inference, and achieves nearly linear inference time within reasonable input audio lengths. It also outperforms existing solutions on output speech audio quality, achieving SDR above 16 over Librimix benchmark.

Learning Epidemiology by Doing The Empirical Implications of a SpatialSIR Model with Behavioral Responses ; We simulate a spatial behavioral model of the diffusion of an infection to understand the role of geographic characteristics the number and distribution of outbreaks, population size, density, and agents' movements. We show that several invariance properties of the SIR model concerning these variables do not hold when agents interact with neighbors in a two dimensional geographical space. Indeed, the spatial model's local interactions generate matching frictions and local herd immunity effects, which play a fundamental role in the infection dynamics. We also show that geographical factors affect how behavioral responses affect the epidemics. We derive relevant implications for estimating the effects of the epidemics and policy interventions that use panel data from several geographical units.

SelfSupervised Noisy Label Learning for SourceFree Unsupervised Domain Adaptation ; It is a strong prerequisite to access source data freely in many existing unsupervised domain adaptation approaches. However, source data is agnostic in many practical scenarios due to the constraints of expensive data transmission and data privacy protection. Usually, the given source domain pretrained model is expected to optimize with only unlabeled target data, which is termed as sourcefree unsupervised domain adaptation. In this paper, we solve this problem from the perspective of noisy label learning, since the given pretrained model can pregenerate noisy label for unlabeled target data via directly network inference. Under this problem modeling, incorporating selfsupervised learning, we propose a novel SelfSupervised Noisy Label Learning method, which can effectively finetune the pretrained model with pregenerated label as well as selfgenerated label on the fly. Extensive experiments had been conducted to validate its effectiveness. Our method can easily achieve stateoftheart results and surpass other methods by a very large margin. Code will be released.

Recurrent Model Predictive Control ; This paper proposes an offline algorithm, called Recurrent Model Predictive Control RMPC, to solve general nonlinear finitehorizon optimal control problems. Unlike traditional Model Predictive Control MPC algorithms, it can make full use of the current computing resources and adaptively select the longest model prediction horizon. Our algorithm employs a recurrent function to approximate the optimal policy, which maps the system states and reference values directly to the control inputs. The number of prediction steps is equal to the number of recurrent cycles of the learned policy function. With an arbitrary initial policy function, the proposed RMPC algorithm can converge to the optimal policy by directly minimizing the designed loss function. We further prove the convergence and optimality of the RMPC algorithm thorough Bellman optimality principle, and demonstrate its generality and efficiency using two numerical examples.

What Doesn't Kill You Makes You Robuster How to Adversarially Train against Data Poisoning ; Data poisoning is a threat model in which a malicious actor tampers with training data to manipulate outcomes at inference time. A variety of defenses against this threat model have been proposed, but each suffers from at least one of the following flaws they are easily overcome by adaptive attacks, they severely reduce testing performance, or they cannot generalize to diverse data poisoning threat models. Adversarial training, and its variants, are currently considered the only empirically strong defense against inferencetime adversarial attacks. In this work, we extend the adversarial training framework to defend against trainingtime data poisoning, including targeted and backdoor attacks. Our method desensitizes networks to the effects of such attacks by creating poisons during training and injecting them into training batches. We show that this defense withstands adaptive attacks, generalizes to diverse threat models, and incurs a better performance tradeoff than previous defenses such as DPSGD or evasion adversarial training.

DataEnhanced Process Models in Process Mining ; Understanding and improving business processes have become important success factors for organizations. Process mining has proven very successful with a variety of methods and techniques, including discovering process models based on event logs. Process mining has traditionally focussed on control flow and timing aspects. However, getting insights about a process is not only based on activities and their orderings, but also on the data generated and manipulated during process executions. Today, almost every process activity generates data; these data do not play the role in process mining that it deserves. This paper introduces a visualization technique for enhancing discovered process models with domain data, thereby allowing databased exploration of processes. Dataenhanced process models aim at supporting domain experts to explore the process, where they can select attributes of interest and observe their influence on the process. The visualization technique is illustrated by the MIMICIV realworld data set on hospitalizations in the US.

Sets of Marginals and PearsonCorrelationbased CHSH Inequalities for a TwoQubit System ; Quantum mass functions QMFs, which are tightly related to decoherence functionals, were introduced by Loeliger and Vontobel IEEE Trans. Inf. Theory, 2017, 2020 as a generalization of probability mass functions toward modeling quantum information processing setups in terms of factor graphs. Simple quantum mass functions SQMFs are a special class of QMFs that do not explicitly model classical random variables. Nevertheless, classical random variables appear implicitly in an SQMF if some marginals of the SQMF satisfy some conditions; variables of the SQMF corresponding to these emerging random variables are called classicable variables. Of particular interest are jointly classicable variables. In this paper we initiate the characterization of the set of marginals given by the collection of jointly classicable variables of a graphical model and compare them with other concepts associated with graphical models like the sets of realizable marginals and the local marginal polytope. In order to further characterize this set of marginals given by the collection of jointly classicable variables, we generalize the CHSH inequality based on the Pearson correlation coefficients, and thereby prove a conjecture proposed by Pozsgay et al. A crucial feature of this inequality is its nonlinearity, which poses difficulties in the proof.

SoundStream An EndtoEnd Neural Audio Codec ; We present SoundStream, a novel neural audio codec that can efficiently compress speech, music and general audio at bitrates normally targeted by speechtailored codecs. SoundStream relies on a model architecture composed by a fully convolutional encoderdecoder network and a residual vector quantizer, which are trained jointly endtoend. Training leverages recent advances in texttospeech and speech enhancement, which combine adversarial and reconstruction losses to allow the generation of highquality audio content from quantized embeddings. By training with structured dropout applied to quantizer layers, a single model can operate across variable bitrates from 3kbps to 18kbps, with a negligible quality loss when compared with models trained at fixed bitrates. In addition, the model is amenable to a low latency implementation, which supports streamable inference and runs in real time on a smartphone CPU. In subjective evaluations using audio at 24kHz sampling rate, SoundStream at 3kbps outperforms Opus at 12kbps and approaches EVS at 9.6kbps. Moreover, we are able to perform joint compression and enhancement either at the encoder or at the decoder side with no additional latency, which we demonstrate through background noise suppression for speech.

Model compression as constrained optimization, with application to neural nets. Part V combining compressions ; Model compression is generally performed by using quantization, lowrank approximation or pruning, for which various algorithms have been researched in recent years. One fundamental question is what types of compression work better for a given model Or even better can we improve by combining compressions in a suitable way We formulate this generally as a problem of optimizing the loss but where the weights are constrained to equal an additive combination of separately compressed parts; and we give an algorithm to learn the corresponding parts' parameters. Experimentally with deep neural nets, we observe that 1 we can find significantly better models in the errorcompression space, indicating that different compression types have complementary benefits, and 2 the best type of combination depends exquisitely on the type of neural net. For example, we can compress ResNets and AlexNet using only 1 bit per weight without error degradation at the cost of adding a few floating point weights. However, VGG nets can be better compressed by combining lowrank with a few floating point weights.

Modeling mechanochemical pattern formation in elastic sheets of biological matter ; Inspired by active shape morphing in developing tissues and biomaterials, we investigate two generic mechanochemical models where the deformations of a thin elastic sheet are driven by, and in turn affect, the concentration gradients of a chemical signal. We develop numerical methods to study the coupled elastic deformations and chemical concentration kinetics, and illustrate with computations the formation of different patterns depending on shell thickness, strength of mechanochemical coupling and diffusivity. In the first model, the sheet curvature governs the production of a contractility inhibitor and depending on the threshold in the coupling, qualitatively different patterns occur. The second model is based on the stressdependent activity of myosin motors, and demonstrates how the concentration distribution patterns of molecular motors are affected by the longrange deformations generated by them. Since the propagation of mechanical deformations is typically faster than chemical kinetics of molecular motors or signaling agents that affect motors, we describe in detail and implement a numerical method based on separation of timescales to effectively simulate such systems. We show that mechanochemical coupling leads to longrange propagation of patterns in disparate systems through elastic instabilities even without the diffusive or advective transport of the chemicals.

DynaDogT A Parametric Animal Model for Synthetic Canine Image Generation ; Synthetic data is becoming increasingly common for training computer vision models for a variety of tasks. Notably, such data has been applied in tasks related to humans such as 3D pose estimation where data is either difficult to create or obtain in realistic settings. Comparatively, there has been less work into synthetic animal data and it's uses for training models. Consequently, we introduce a parametric canine model, DynaDogT, for generating synthetic canine images and data which we use for a common computer vision task, binary segmentation, which would otherwise be difficult due to the lack of available data.

A Markov Game Model for AIbased Cyber Security Attack Mitigation ; The new generation of cyber threats leverages advanced AIaided methods, which make them capable to launch multistage, dynamic, and effective attacks. Current cyberdefense systems encounter various challenges to defend against such new and emerging threats. Modeling AIaided threats through game theory models can help the defender to select optimal strategies against the attacks and make wise decisions to mitigate the attack's impact. This paper first explores the current stateoftheart in the new generation of threats in which AI techniques such as deep neural network is used for the attacker and discusses further challenges. We propose a Markovian dynamic game that can evaluate the efficiency of defensive methods against the AIaided attacker under a cloudbased system in which the attacker utilizes an AI technique to launch an advanced attack by finding the shortest attack path. We use the CVSS metrics to quantify the values of this zerosum game model for decisionmaking.

An ultraviolet completion for the Scotogenic model ; The Scotogenic model is an economical scenario that generates neutrino masses at the 1loop level and includes a dark matter candidate. This is achieved by means of an adhoc mathbbZ2 symmetry, which forbids the treelevel generation of neutrino masses and stabilizes the lightest mathbbZ2odd state. Neutrino masses are also suppressed by a quartic coupling, usually denoted by lambda5. While the smallness of this parameter is natural, it is not explained in the context of the Scotogenic model. We construct an ultraviolet completion of the Scotogenic model that provides a natural explanation for the smallness of the lambda5 parameter and induces the mathbbZ2 parity as the lowenergy remnant of a global rm U1 symmetry at high energies. The lowenergy spectrum contains, besides the usual Scotogenic states, a massive scalar and a massless Goldstone boson, hence leading to novel phenomenological predictions in flavor observables, dark matter physics and colliders.

DoubleRobust TwoWayFixedEffects Regression For Panel Data ; We propose a new estimator for the average causal effects of a binary treatment with panel data in settings with general treatment patterns. Our approach augments the twowayfixedeffects specification with the unitspecific weights that arise from a model for the assignment mechanism. We show how to construct these weights in various settings, including situations where units opt into the treatment sequentially. The resulting estimator converges to an average over units and time treatment effect under the correct specification of the assignment model. We show that our estimator is more robust than the conventional twoway estimator it remains consistent if either the assignment mechanism or the twoway regression model is correctly specified and performs better than the twowayfixedeffect estimator if both are locally misspecified. This strong double robustness property quantifies the benefits from modeling the assignment process and motivates using our estimator in practice.

Creating Powerful and Interpretable Models with Regression Networks ; As the discipline has evolved, research in machine learning has been focused more and more on creating more powerful neural networks, without regard for the interpretability of these networks. Such blackbox models yield stateoftheart results, but we cannot understand why they make a particular decision or prediction. Sometimes this is acceptable, but often it is not. We propose a novel architecture, Regression Networks, which combines the power of neural networks with the understandability of regression analysis. While some methods for combining these exist in the literature, our architecture generalizes these approaches by taking interactions into account, offering the power of a dense neural network without forsaking interpretability. We demonstrate that the models exceed the stateoftheart performance of interpretable models on several benchmark datasets, matching the power of a dense neural network. Finally, we discuss how these techniques can be generalized to other neural architectures, such as convolutional and recurrent neural networks.

Completeness of Bethe Ansatz by Sklyanin SOV for Cyclic Representations of Integrable Quantum Models ; In 1 an integrable quantum model was introduced and a class of its cyclic representations was proven to define lattice regularizations of the SineGordon model. Here, we analyze general cyclic representations of this integrable quantum model by extending the spectrum construction introduced in 2 in the framework of the Separation of Variables SOV of Sklyanin. We show that as in 1 also for general representations, the transfer matrix spectrum eigenvalues and eigenstates is completely characterized in terms of polynomial solutions of an associated functional Baxter equation. Moreover, we prove that the method here developed has two fundamental builtin features i the completeness of the set of the transfer matrix eigenstates constructed from the solutions of the associated Bethe ansatz equations, ii the existence and complete characterization of the Baxter Qoperator.

Finitelattice form factors in freefermion models ; We consider the general mathbbZ2symmetric freefermion model on the finite periodic lattice, which includes as special cases the Ising model on the square and triangular lattices and mathbbZnsymmetric BBS tau2model with n2. Translating Kaufman's fermionic approach to diagonalization of Isinglike transfer matrices into the language of Grassmann integrals, we determine the transfer matrix eigenvectors and observe that they coincide with the eigenvectors of a square lattice Ising transfer matrix. This allows to find exact finitelattice form factors of spin operators for the statistical model and the associated finitelength quantum chains, of which the most general is equivalent to the XY chain in a transverse field.

Permanence of a general discretetime twospeciesinteraction model with nonmonotonic per capita growth rates ; Combined with all densitydependent factors, the per capita growth rate of a species may be nonmonotonic. One important consequence is that species may suffer from weak Allee effects or strong Allee effects. In this paper, we study the permanence of a discretetime twospeciesinteraction model with nonmonotonic per capita growth rates for the first time. By using the average Lyapunov functions and extending the ecological concept of the relative nonlinearity, we find a simple sufficient condition for guaranteeing the permanence of systems that can model complicated twospecies interactions. The extended relative nonlinearity allows us to fully characterize the effects of nonlinearities in the per capita growth functions with nonmonotonicity. These results are illustrated with specific two species competition and predatorprey models of generic forms with nonmonotone per capita growth rates.

Numerical analysis of cosmological models for accelerating Universe in Poincare gauge theory of gravity ; Homogeneous isotropic models with two torsion functions built in the framework of the Poincare gauge theory of gravity based on general expression of gravitational Lagrangian by certain restrictions on indefinite parameters are analyzed numerically. Special points of cosmological solutions at asymptotics and conditions of their stability in dependence of indefinite parameters are found. Procedure of numerical integration of the system of gravitational equations at asymptotics is considered. Numerical solution for accelerating Universe without dark energy and dark matter is obtained. It is shown that by certain restrictions on indefinite parameters obtained cosmological solutions are in agreement with SNe Ia observational data and Big Bang Nucleosynthesis predictions. Statefinder diagnostics is discussed in order to compare considered cosmological model with other models.

Evolution of nonGaussianity in multiscalar field models ; We study the evolution of nonGaussianity in multiplefield inflationary models, focusing on three fundamental questions a How is the sign and peak magnitude of the nonlinearity parameter fNL related to generic features in the inflationary potential b How sensitive is fNL to the process by which an adiabatic limit is reached, where the curvature perturbation becomes conserved c For a given model, what is the appropriate tool analytic or numerical to calculate fNL at the adiabatic limit We summarise recent results obtained by the authors and further elucidate them by considering an inflection point model.

Applying generalized Pade approximants in analytic QCD models ; A method of resummation of truncated perturbation series, related to diagonal Pad'e approximants but giving results independent of the renormalization scale, was developed more than ten years ago by us with a view of applying it in perturbative QCD. We now apply this method in analytic QCD models, i.e., models where the running coupling has no unphysical singularities, and we show that the method has attractive features such as a rapid convergence. The method can be regarded as a generalization of the scalesetting methods of Stevenson, Grunberg, and BrodskyLepageMackenzie. The method involves the fixing of various scales and weight coefficients via an auxiliary construction of diagonal Pad'e approximant. In lowenergy QCD observables, some of these scales become sometimes low at high order, which prevents the method from being effective in perturbative QCD where the coupling has unphysical singularities at low spacelike momenta. There are no such problems in analytic QCD.

Type IIB Supersymmetric Flux Vacua ; On the Type IIB toroidal T6 orientifolds with generic flux compactifications, we conjecture that in generic supsersymmetric Minkowski vacua, at least one of the flux contributions to the sevenbrane and D3brane tadpoles is positive if the moduli are stabilized properly, and then the tadpole cancellation conditions can not be relaxed. To study the supsersymmetric Minkowski flux vacua, we simplify the fluxes reasonably and discuss the corresponding superpotential. We show that we can not have simultaneously the positive real parts of all the moduli and the negativezero flux contributions to all the sevenbrane and D3brane tadpoles. Therefore, we can not construct realistic flux models with the relaxed tadpole cancellation conditions. When studying the supsersymmetric AdS vacua, we obtain flux models with the sevenbrane and D3brane tadpole cancellation conditions relaxed elegantly, and we present a semirealistic PatiSalam model as well as its particle spectrum. The lifting from the AdS vacua to the MinkowskidS vacua remains a great challenge in flux model buildings on toroidal orientifolds.

Powerlaw entropy corrected new holographic scalar field models of dark energy with modified IRcutoff ; In this work, the PLECHDE model with GrandaOliveros GO IRcutoff is studied. The evolution of dark energy density, deceleration and EoS parameters are calculated. I demonstrate that under a condition, our universe can accelerate near the phantom barrier at present time. We calculate these parameters also in PLECHDE at Ricci scale, when when alpha2 and beta1, and at last a comparison between Ricci scale, GO cutoff and noncorrected HDE without matter field with GO cutoff is done. The correspondence between this model and some scalar field of dark energy models is established. By this method, the evolutionary treatment of kinetic energy and potential for quintessence, tachyon, Kessence and dilaton fields, are obtained. I show that the results has a good compatibility with previous work in the limiting case of flat, dark dominated and non corrected holographic dark energy.

LRS BianchiI Anisotropic Cosmological Model with Dominance of Dark Energy ; The present study deals with spatially homogeneous and anisotropic locally rotationally symmetric LRS Bianchi type I cosmological model with dominance of dark energy. To get the deterministic model of Universe, we assume that the shear scalar sigma in the model is proportional to expansion scalar theta. This condition leads to ABn, where A,;B are metric potential and n is positive constant. It has been found that the anisotropic distribution of dark energy leads to the present accelerated expansion of Universe. The physical behavior of the Universe has been discussed in detail.

Inverse spin glass and related maximum entropy problems ; If we have a system of binary variables and we measure the pairwise correlations among these variables, then the least structured or maximum entropy model for their joint distribution is an Ising model with pairwise interactions among the spins. Here we consider inhomogeneous systems in which we constrain for example not the full matrix of correlations, but only the distribution from which these correlations are drawn. In this sense, what we have constructed is an inverse spin glass rather than choosing coupling constants at random from a distribution and calculating correlations, we choose the correlations from a distribution and infer the coupling constants. We argue that such models generate a block structure in the space of couplings, which provides an explicit solution of the inverse problem. This allows us to generate a phase diagram in the space of measurable moments of the distribution of correlations. We expect that these ideas will be most useful in building models for systems that are nonequilibrium statistical mechanics problems, such as networks of real neurons.

An Emergent Universe with Dark Sector Fields in a Chiral Cosmological Model ; We consider the emergent universe scenario supported by a chiral cosmological model with two interacting dark sector fields phantom and canonical. We investigate the general properties of the evolution of the kinetic and potential energies as well as the development of the equation of state with time. We present three models based on asymptotic solutions and investigate the phantom part of the potential and chiral metric components. The exact solution corresponding to a global emergent universe scenario, starting from the infinite past and evolving to the infinite future, has been obtained for the first time for a chiral cosmological model. The behavior of the chiral metric components responsible for the kinetic interaction between the phantom and canonical scalar fields has been analyzed as well.