text
stringlengths 62
2.94k
|
|---|
A New Distributed Method for Training Generative Adversarial Networks ; Generative adversarial networks GANs are emerging machine learning models for generating synthesized data similar to real data by jointly training a generator and a discriminator. In many applications, data and computational resources are distributed over many devices, so centralized computation with all data in one location is infeasible due to privacy andor communication constraints. This paper proposes a new framework for training GANs in a distributed fashion Each device computes a local discriminator using local data; a single server aggregates their results and computes a global GAN. Specifically, in each iteration, the server sends the global GAN to the devices, which then update their local discriminators; the devices send their results to the server, which then computes their average as the global discriminator and updates the global generator accordingly. Two different update schedules are designed with different levels of parallelism between the devices and the server. Numerical results obtained using three popular datasets demonstrate that the proposed framework can outperform a stateoftheart framework in terms of convergence speed. |
OverParameterization and Generalization in Audio Classification ; Convolutional Neural Networks CNNs have been dominating classification tasks in various domains, such as machine vision, machine listening, and natural language processing. In machine listening, while generally exhibiting very good generalization capabilities, CNNs are sensitive to the specific audio recording device used, which has been recognized as a substantial problem in the acoustic scene classification DCASE community. In this study, we investigate the relationship between overparameterization of acoustic scene classification models, and their resulting generalization abilities. Specifically, we test scaling CNNs in width and depth, under different conditions. Our results indicate that increasing width improves generalization to unseen devices, even without an increase in the number of parameters. |
Improve Learning from Crowds via Generative Augmentation ; Crowdsourcing provides an efficient label collection schema for supervised machine learning. However, to control annotation cost, each instance in the crowdsourced data is typically annotated by a small number of annotators. This creates a sparsity issue and limits the quality of machine learning models trained on such data. In this paper, we study how to handle sparsity in crowdsourced data using data augmentation. Specifically, we propose to directly learn a classifier by augmenting the raw sparse annotations. We implement two principles of highquality augmentation using Generative Adversarial Networks 1 the generated annotations should follow the distribution of authentic ones, which is measured by a discriminator; 2 the generated annotations should have high mutual information with the groundtruth labels, which is measured by an auxiliary network. Extensive experiments and comparisons against an array of stateoftheart learning from crowds methods on three realworld datasets proved the effectiveness of our data augmentation framework. It shows the potential of our algorithm for lowbudget crowdsourcing in general. |
Linear growth of the entanglement entropy for quadratic Hamiltonians and arbitrary initial states ; We prove that the entanglement entropy of any pure initial state of a bipartite bosonic quantum system grows linearly in time with respect to the dynamics induced by any unstable quadratic Hamiltonian. The growth rate does not depend on the initial state and is equal to the sum of certain Lyapunov exponents of the corresponding classical dynamics. This paper generalizes the findings of Bianchi et al., JHEP 2018, 25 2018, which proves the same result in the special case of Gaussian initial states. Our proof is based on a recent generalization of the strong subadditivity of the von Neumann entropy for bosonic quantum systems De Palma et al., arXiv2105.05627. This technique allows us to extend our result to generic mixed initial states, with the squashed entanglement providing the right generalization of the entanglement entropy. We discuss several applications of our results to physical systems with weakly interacting Hamiltonians and periodically driven quantum systems, including certain quantum field theory models. |
Statefinder and Om Diagnostics for Interacting New Holographic Dark Energy Model and Generalized Second Law of Thermodynamics ; In this work, we have considered that the flat FRW universe is filled with the mixture of dark matter and the new holographic dark energy. If there is an interaction, we have investigated the natures of deceleration parameter, statefinder and Om diagnostics. We have examined the validity of the first and generalized second laws of thermodynamics under these interactions on the event as well as apparent horizon. It has been observed that the first law is violated on the event horizon. However, the generalized second law is valid throughout the evolution of the universe enveloped by the apparent horizon. When the event horizon is considered as the enveloping horizon, the generalized second law is found to break down excepting at late stage of the universe. |
Transition state theory a generalization to nonequilibrium systems with powerlaw distributions ; Transition state theory TST is generalized for the nonequilibrium system with powerlaw distributions. The stochastic dynamics that gives rise to the powerlaw distributions for the reaction coordinate and momentum is modeled by the Langevin equations and corresponding FokkerPlanck equations. It is assumed that the system far away from equilibrium has not to relax to a thermal equilibrium state with BoltzmannGibbs distribution, but asymptotically approaches to a nonequilibrium stationarystate with powerlaw distributions. Thus, we obtain a generalization of TST rates to nonequilibrium systems with powerlaw distributions. Furthermore, we derive the generalized TST rate constants for onedimension and ndimension Hamiltonian systems away from equilibrium, and receive a generalized Arrhenius rate for the system with powerlaw distributions. |
A framework for dynamical generation of flavor mixing ; We present a dynamical mechanism a la NambuJonaLasinio for the generation of masses and mixing for two interacting fermion fields. The analysis is carried out in the framework introduced long ago by Umezawa et al., in which mass generation is achieved via inequivalent representations, and that we generalize to the case of two generations. The method allows a clear identification of the vacuum structure for each physical phase, confirming previous results about the distinct physical nature of the vacuum for fields with definite mass and fields with definite flavor. Implications for the leptonic sector of the Standard Model are briefly discussed. |
Properties of Answer Set Programming with Convex Generalized Atoms ; In recent years, Answer Set Programming ASP, logic programming under the stable model or answer set semantics, has seen several extensions by generalizing the notion of an atom in these programs be it aggregate atoms, HEX atoms, generalized quantifiers, or abstract constraints, the idea is to have more complicated satisfaction patterns in the lattice of Herbrand interpretations than traditional, simple atoms. In this paper we refer to any of these constructs as generalized atoms. Several semantics with differing characteristics have been proposed for these extensions, rendering the big picture somewhat blurry. In this paper, we analyze the class of programs that have convex generalized atoms originally proposed by Liu and Truszczynski in 10 in rule bodies and show that for this class many of the proposed semantics coincide. This is an interesting result, since recently it has been shown that this class is the precise complexity boundary for the FLP semantics. We investigate whether similar results also hold for other semantics, and discuss the implications of our findings. |
Even more generic solution construction in ValuationBased Systems ; Valuation algebras abstract a large number of formalisms for automated reasoning and enable the definition of generic inference procedures. Many of these formalisms provide some notions of solutions. Typical examples are satisfying assignments in constraint systems, models in logics or solutions to linear equation systems. Recently, formal requirements for the presence of solutions and a generic algorithm for solution construction based on the results of a previously executed inference scheme have been proposed in the literature. Unfortunately, the formalization of Pouly and Kohlas relies on a theorem for which we provide a counter example. In spite of that, the mainline of the theory described is correct, although some of the necessary conditions to apply some of the algorithms have to be revised. To fix the theory, we generalize some of their definitions and provide correct sufficient conditions for the algorithms. As a result, we get a more general and corrected version of the already existing theory. |
Multidomain Neural Network Language Generation for Spoken Dialogue Systems ; Moving from limiteddomain natural language generation NLG to open domain is difficult because the number of semantic input combinations grows exponentially with the number of domains. Therefore, it is important to leverage existing resources and exploit similarities between domains to facilitate domain adaptation. In this paper, we propose a procedure to train multidomain, Recurrent Neural Networkbased RNN language generators via multiple adaptation steps. In this procedure, a model is first trained on counterfeited data synthesised from an outofdomain dataset, and then fine tuned on a small set of indomain utterances with a discriminative objective function. Corpusbased evaluation results show that the proposed procedure can achieve competitive performance in terms of BLEU score and slot error rate while significantly reducing the data needed to train generators in new, unseen domains. In subjective testing, human judges confirm that the procedure greatly improves generator performance when only a small amount of data is available in the domain. |
Generalised relyguarantee concurrency An algebraic foundation ; The relyguarantee technique allows one to reason compositionally about concurrent programs. To handle interference the technique makes use of rely and guarantee conditions, both of which are binary relations on states. A rely condition is an assumption that the environment performs only atomic steps satisfying the rely relation and a guarantee is a commitment that every atomic step the program makes satisfies the guarantee relation. In order to investigate relyguarantee reasoning more generally, in this paper we allow interference to be represented by a process rather than a relation and hence derive more general relyguarantee laws. The paper makes use of a weak conjunction operator between processes, which generalises a guarantee relation to a guarantee process, and introduces a rely quotient operator, which generalises a rely relation to a process. The paper focuses on the algebraic properties of the general relyguarantee theory. The Jonesstyle relyguarantee theory can be interpreted as a model of the general algebraic theory and hence the general laws presented here hold for that theory. |
Texture Networks Feedforward Synthesis of Textures and Stylized Images ; Gatys et al. recently demonstrated that deep networks can generate beautiful textures and stylized images from a single texture example. However, their methods requires a slow and memoryconsuming optimization process. We propose here an alternative approach that moves the computational burden to a learning stage. Given a single example of a texture, our approach trains compact feedforward convolutional networks to generate multiple samples of the same texture of arbitrary size and to transfer artistic style from a given image to any other image. The resulting networks are remarkably lightweight and can generate textures of quality comparable to Gatysetal., but hundreds of times faster. More generally, our approach highlights the power and flexibility of generative feedforward models trained with complex and expressive loss functions. |
Towards Optimal Energy Management of Microgrids with a Realistic Model ; This work considers energy management in a gridconnected microgrid which consists of multiple conventional generators CGs, renewable generators RGs and energy storage systems ESSs. A twostage optimization approach is presented to schedule the power generation, aimed at minimizing the longterm average operating cost subject to operational and service constraints. The first stage of optimization determines hourly unit commitment of the CGs via a dayahead scheduling, and the second stage performs economic dispatch of the CGs, ESSs and energy trading via an hourahead scheduling. The combined solution meets the need of handling large uncertainties in the load demand and renewable generation, and provides an efficient solution under limited computational resource which meets both shortterm and longterm qualityofservice requirements. The performance of the proposed strategy is evaluated by simulations based on real load demand and renewable generation data. |
Generation and Evaluation of SpaceTime Trajectories of Photovoltaic Power ; In the probabilistic energy forecasting literature, emphasis is mainly placed on deriving marginal predictive densities for which each random variable is dealt with individually. Such marginals description is sufficient for power systems related operational problems if and only if optimal decisions are to be made for each leadtime and each location independently of each other. However, many of these operational processes are temporally and spatially coupled, while uncertainty in photovoltaic PV generation is strongly dependent in time and in space. This issue is addressed here by analysing and capturing spatiotemporal dependencies in PV generation. Multivariate predictive distributions are modelled and spacetime trajectories describing the potential evolution of forecast errors through successive leadtimes and locations are generated. Discrimination ability of the relevant scoring rules on performance assessment of spacetime trajectories of PV generation is also studied. Finally, the advantage of taking into account spacetime correlations over probabilistic and point forecasts is investigated. The empirical investigation is based on the solar PV dataset of the Global Energy Forecasting Competition GEFCom 2014. |
Latent Predictor Networks for Code Generation ; Many language generation tasks require the production of text conditioned on both structured and unstructured inputs. We present a novel neural network architecture which generates an output sequence conditioned on an arbitrary number of input functions. Crucially, our approach allows both the choice of conditioning context and the granularity of generation, for example characters or tokens, to be marginalised, thus permitting scalable and effective training. Using this framework, we address the problem of generating programming code from a mixed natural language and structured specification. We create two new data sets for this paradigm derived from the collectible trading card games Magic the Gathering and Hearthstone. On these, and a third preexisting corpus, we demonstrate that marginalising multiple predictors allows our model to outperform strong benchmarks. |
The Beta Generalized MarshallOlkinG Family of Distributions ; In this paper we propose a new family of distribution considering Generalized MarshalOlkin distribution as the base line distribution in the BetaG family of Construction. The new family includes BetaG Eugene et al. 2002 and Jones, 2004 and Jayakumar and Mathew, 2008 families as particular cases. Probability density function pdf and the cumulative distribution function cdf are expressed as mixture of the MarshalOlkin Marshal and Olkin, 1997 distribution. Series expansions of pdf of the order statistics are also obtained. Moments, moment generating function, R'enyi entropies, quantile power series, random sample generation and asymptotes are also investigated. Parameter estimation by method of maximum likelihood and method of moment are also presented. Finally proposed model is compared to the Generalized MarshallOlkin Kumaraswamy extended family Handique and Chakraborty, 2015 by considering three data fitting examples with real life data sets. |
Generalized BransDicke inflation with a quartic potential ; Within the framework of BransDicke gravity, we investigate inflation with the quartic potential, lambdavarphi44, in the presence of generalized BransDicke parameter omegarm GBDvarphi. We obtain the inflationary observables containing the scalar spectral index, the tensortoscalar ratio, the running of the scalar spectral index and the equilateral nonGaussianity parameter in terms of general form of the potential Uvarphi and omegarm GBDvarphi. For the quartic potential, our results show that the predictions of the model are in well agreement with the Planck 2015 data for the generalized BransDicke parameters omegarm GBDvarphiomega0varphin and omega0ebvarphi. This is in contrast with both the Einstein and standard BransDicke gravity, in which the result of quartic potential is disfavored by the Planck data. |
SelfAveraging Expectation Propagation ; We investigate the problem of approximate Bayesian inference for a general class of observation models by means of the expectation propagation EP framework for large systems under some statistical assumptions. Our approach tries to overcome the numerical bottleneck of EP caused by the inversion of large matrices. Assuming that the measurement matrices are realizations of specific types of ensembles we use the concept of freeness from random matrix theory to show that the EP cavity variances exhibit an asymptotic selfaveraging property. They can be precomputed using specific generating functions, i.e. the R andor Stransforms in free probability, which do not require matrix inversions. Our approach extends the framework of generalized approximate message passing assumes zeromean iid entries of the measurement matrix to a general class of random matrix ensembles. The generalization is via a simple formulation of the R andor Stransforms of the limiting eigenvalue distribution of the Gramian of the measurement matrix. We demonstrate the performance of our approach on a signal recovery problem of nonlinear compressed sensing and compare it with that of EP. |
Generating Realistic Geology Conditioned on Physical Measurements with Generative Adversarial Networks ; An important problem in geostatistics is to build models of the subsurface of the Earth given physical measurements at sparse spatial locations. Typically, this is done using spatial interpolation methods or by reproducing patterns from a reference image. However, these algorithms fail to produce realistic patterns and do not exhibit the wide range of uncertainty inherent in the prediction of geology. In this paper, we show how semantic inpainting with Generative Adversarial Networks can be used to generate varied realizations of geology which honor physical measurements while matching the expected geological patterns. In contrast to other algorithms, our method scales well with the number of data points and mimics a distribution of patterns as opposed to a single pattern or image. The generated conditional samples are state of the art. |
PseudoRecursal Solving the Catastrophic Forgetting Problem in Deep Neural Networks ; In general, neural networks are not currently capable of learning tasks in a sequential fashion. When a novel, unrelated task is learnt by a neural network, it substantially forgets how to solve previously learnt tasks. One of the original solutions to this problem is pseudorehearsal, which involves learning the new task while rehearsing generated items representative of the previous tasks. This is very effective for simple tasks. However, pseudorehearsal has not yet been successfully applied to very complex tasks because in these tasks it is difficult to generate representative items. We accomplish pseudorehearsal by using a Generative Adversarial Network to generate items so that our deep network can learn to sequentially classify the CIFAR10, SVHN and MNIST datasets. After training on all tasks, our network loses only 1.67 absolute accuracy on CIFAR10 and gains 0.24 absolute accuracy on SVHN. Our model's performance is a substantial improvement compared to the current state of the art solution. |
GeometryBased Data Generation ; Many generative models attempt to replicate the density of their input data. However, this approach is often undesirable, since data density is highly affected by sampling biases, noise, and artifacts. We propose a method called SUGAR Synthesis Using Geometrically Aligned Randomwalks that uses a diffusion process to learn a manifold geometry from the data. Then, it generates new points evenly along the manifold by pulling randomly generated points into its intrinsic structure using a diffusion kernel. SUGAR equalizes the density along the manifold by selectively generating points in sparse areas of the manifold. We demonstrate how the approach corrects sampling biases and artifacts, while also revealing intrinsic patterns e.g. progression and relations in the data. The method is applicable for correcting missing data, finding hypothetical data points, and learning relationships between data features. |
Learning Hyperedge Replacement Grammars for Graph Generation ; The discovery and analysis of network patterns are central to the scientific enterprise. In the present work, we developed and evaluated a new approach that learns the building blocks of graphs that can be used to understand and generate new realistic graphs. Our key insight is that a graph's clique tree encodes robust and precise information. We show that a Hyperedge Replacement Grammar HRG can be extracted from the clique tree, and we develop a fixedsize graph generation algorithm that can be used to produce new graphs of a specified size. In experiments on large realworld graphs, we show that graphs generated from the HRG approach exhibit a diverse range of properties that are similar to those found in the original networks. In addition to graph properties like degree or eigenvector centrality, what a graph looks like ultimately depends on small details in local graph substructures that are difficult to define at a global level. We show that the HRG model can also preserve these local substructures when generating new graphs. |
Constrained Image Generation Using Binarized Neural Networks with Decision Procedures ; We consider the problem of binary image generation with given properties. This problem arises in a number of practical applications, including generation of artificial porous medium for an electrode of lithiumion batteries, for composed materials, etc. A generated image represents a porous medium and, as such, it is subject to two sets of constraints topological constraints on the structure and process constraints on the physical process over this structure. To perform image generation we need to define a mapping from a porous medium to its physical process parameters. For a given geometry of a porous medium, this mapping can be done by solving a partial differential equation PDE. However, embedding a PDE solver into the search procedure is computationally expensive. We use a binarized neural network to approximate a PDE solver. This allows us to encode the entire problem as a logical formula. Our main contribution is that, for the first time, we show that this problem can be tackled using decision procedures. Our experiments show that our model is able to produce random constrained images that satisfy both topological and process constraints. |
Generalized modes in Bayesian inverse problems ; Uncertainty quantification requires efficient summarization of high or even infinitedimensional i.e., nonparametric distributions based on, e.g., suitable point estimates modes for posterior distributions arising from modelspecific prior distributions. In this work, we consider nonparametric modes and MAP estimates for priors that do not admit continuous densities, for which previous approaches based on small ball probabilities fail. We propose a novel definition of generalized modes based on the concept of approximating sequences, which reduce to the classical mode in certain situations that include Gaussian priors but also exist for a more general class of priors. The latter includes the case of priors that impose strict bounds on the admissible parameters and in particular of uniform priors. For uniform priors defined by random series with uniformly distributed coefficients, we show that generalized MAP estimates but not classical MAP estimates can be characterized as minimizers of a suitable functional that plays the role of a generalized OnsagerMachlup functional. This is then used to show consistency of nonlinear Bayesian inverse problems with uniform priors and Gaussian noise. |
Best sources forward domain generalization through sourcespecific nets ; A long standing problem in visual object categorization is the ability of algorithms to generalize across different testing conditions. The problem has been formalized as a covariate shift among the probability distributions generating the training data source and the test data target and several domain adaptation methods have been proposed to address this issue. While these approaches have considered the single sourcesingle target scenario, it is plausible to have multiple sources and require adaptation to any possible target domain. This last scenario, named Domain Generalization DG, is the focus of our work. Differently from previous DG methods which learn domain invariant representations from source data, we design a deep network with multiple domainspecific classifiers, each associated to a source domain. At test time we estimate the probabilities that a target sample belongs to each source domain and exploit them to optimally fuse the classifiers predictions. To further improve the generalization ability of our model, we also introduced a domain agnostic component supporting the final classifier. Experiments on two public benchmarks demonstrate the power of our approach. |
The Neural Painter MultiTurn Image Generation ; In this work we combine two research threads from Vision Graphics and Natural Language Processing to formulate an image generation task conditioned on attributes in a multiturn setting. By multiturn, we mean the image is generated in a series of steps of userspecified conditioning information. Our proposed approach is practically useful and offers insights into neural interpretability. We introduce a framework that includes a novel training algorithm as well as model improvements built for the multiturn setting. We demonstrate that this framework generates a sequence of images that match the given conditioning information and that this task is useful for more detailed benchmarking and analysis of conditional image generation methods. |
A precise extragalactic test of General Relativity ; Einstein's theory of gravity, General Relativity, has been precisely tested on Solar System scales, but the longrange nature of gravity is still poorly constrained. The nearby strong gravitational lens, ESO 325G004, provides a laboratory to probe the weakfield regime of gravity and measure the spatial curvature generated per unit mass, gamma. By reconstructing the observed light profile of the lensed arcs and the observed spatially resolved stellar kinematics with a single selfconsistent model, we conclude that gamma 0.97 pm 0.09 at 68 confidence. Our result is consistent with the prediction of 1 from General Relativity and provides a strong extragalactic constraint on the weakfield metric of gravity. |
Neuralnetinduced Gaussian process regression for function approximation and PDE solution ; Neuralnetinduced Gaussian process NNGP regression inherits both the high expressivity of deep neural networks deep NNs as well as the uncertainty quantification property of Gaussian processes GPs. We generalize the current NNGP to first include a larger number of hyperparameters and subsequently train the model by maximum likelihood estimation. Unlike previous works on NNGP that targeted classification, here we apply the generalized NNGP to function approximation and to solving partial differential equations PDEs. Specifically, we develop an analytical iteration formula to compute the covariance function of GP induced by deep NN with an errorfunction nonlinearity. We compare the performance of the generalized NNGP for function approximations and PDE solutions with those of GPs and fullyconnected NNs. We observe that for smooth functions the generalized NNGP can yield the same order of accuracy with GP, while both NNGP and GP outperform deep NN. For nonsmooth functions, the generalized NNGP is superior to GP and comparable or superior to deep NN. |
Generate the corresponding Image from Text Description using Modified GANCLS Algorithm ; Synthesizing images or texts automatically is a useful research area in the artificial intelligence nowadays. Generative adversarial networks GANs, which are proposed by Goodfellow in 2014, make this task to be done more efficiently by using deep neural networks. We consider generating corresponding images from an input text description using a GAN. In this paper, we analyze the GANCLS algorithm, which is a kind of advanced method of GAN proposed by Scott Reed in 2016. First, we find the problem with this algorithm through inference. Then we correct the GANCLS algorithm according to the inference by modifying the objective function of the model. Finally, we do the experiments on the Oxford102 dataset and the CUB dataset. As a result, our modified algorithm can generate images which are more plausible than the GANCLS algorithm in some cases. Also, some of the generated images match the input texts better. |
Convergence Problems with Generative Adversarial Networks GANs ; Generative adversarial networks GANs are a novel approach to generative modelling, a task whose goal it is to learn a distribution of real data points. They have often proved difficult to train GANs are unlike many techniques in machine learning, in that they are best described as a twoplayer game between a discriminator and generator. This has yielded both unreliability in the training process, and a general lack of understanding as to how GANs converge, and if so, to what. The purpose of this dissertation is to provide an account of the theory of GANs suitable for the mathematician, highlighting both positive and negative results. This involves identifying the problems when training GANs, and how topological and gametheoretic perspectives of GANs have contributed to our understanding and improved our techniques in recent years. |
Crossview image synthesis using geometryguided conditional GANs ; We address the problem of generating images across two drastically different views, namely ground street and aerial overhead views. Image synthesis by itself is a very challenging computer vision task and is even more so when generation is conditioned on an image in another view. Due the difference in viewpoints, there is small overlapping field of view and little common content between these two views. Here, we try to preserve the pixel information between the views so that the generated image is a realistic representation of cross view input image. For this, we propose to use homography as a guide to map the images between the views based on the common field of view to preserve the details in the input image. We then use generative adversarial networks to inpaint the missing regions in the transformed image and add realism to it. Our exhaustive evaluation and model comparison demonstrate that utilizing geometry constraints adds fine details to the generated images and can be a better approach for cross view image synthesis than purely pixel based synthesis methods. |
On numerical nonvanishing for generalized log canonical pairs ; The nonvanishing conjecture for projective log canonical pairs plays a key role in the minimal model program of higher dimensional algebraic geometry. The numerical nonvanishing conjecture considered in this paper is a weaker version of the usual nonvanishing conjecture, but valid in the more general setting of generalized log canonical pairs. We confirm it in dimension two. Under some necessary conditions we obtain effective versions of numerical nonvanishing for surfaces. Several applications are also discussed. In higher dimensions, we mainly consider the conjecture for generalized klt pairs X, BmathbfM, and reduce it to lower dimensions when KXmathbfMX is not pseudoeffective. Up to scaling the nef part, we prove the numerical nonvanishing for pseudoeffective generalized lc threefolds with rational singularities. |
State of the art POWHEG generators for top mass measurements at the LHC ; We study the theoretical uncertainties in the determination of the topquark mass using nexttoleadingorder NLO generators, that describe the topquark decay at different levels of accuracy, interfaced to parton showers PS. Specifically we consider one generator that implements NLO corrections only in the production dynamics, one that also takes them into account in the topquark decay in the narrow width approximation NWA and one that implements them exactly, including finitewidth and interference effects. We aim at assessing the errors in topmass determinations of purely theoretical origin. We do so by measuring relative peak position shifts of Wbjet mass distributions. Besides the theoretical errors due to the use of less accurate NLOPS generators, we also explore uncertainties related to shower and modelling of nonperturbative effects by comparing the results obtained by interfacing our generators to both Pythia and Herwig shower Monte Carlos SMCs. |
Generalized Solitary Waves in a FiniteDifference Kortewegde Vries Equation ; Generalized solitary waves with exponentially small nondecaying far field oscillations have been studied in a range of singularlyperturbed differential equations, including higherorder Kortewegde Vries KdV equations. Many of these studies used exponential asymptotics to compute the behaviour of the oscillations, revealing that they appear in the solution as special curves known as Stokes lines are crossed. Recent studies have identified similar behaviour in solutions to difference equations. Motivated by these studies, the seventhorder KdV and a hierarchy of higherorder KdV equations are investigated, identifying conditions which produce generalized solitary wave solutions. These results form a foundation for the study of infiniteorder differential equations, which are used as a model for studying lattice equations. Finally, a lattice KdV equation is generated using finitedifference discretization, in which a lattice generalized solitary wave solution is found. |
Generative Adversarial Network for Medical Images MIGAN ; Deep learning algorithms produces stateoftheart results for different machine learning and computer vision tasks. To perform well on a given task, these algorithms require large dataset for training. However, deep learning algorithms lack generalization and suffer from overfitting whenever trained on small dataset, especially when one is dealing with medical images. For supervised image analysis in medical imaging, having image data along with their corresponding annotated groundtruths is costly as well as time consuming since annotations of the data is done by medical experts manually. In this paper, we propose a new Generative Adversarial Network for Medical Imaging MIGAN. The MIGAN generates synthetic medical images and their segmented masks, which can then be used for the application of supervised analysis of medical images. Particularly, we present MIGAN for synthesis of retinal images. The proposed method generates precise segmented images better than the existing techniques. The proposed model achieves a dice coefficient of 0.837 on STARE dataset and 0.832 on DRIVE dataset which is stateoftheart performance on both the datasets. |
Hilltop inflation and generation of helical magnetic field ; Primordial magnetic field generated in the inflationary era can act as a viable source for the present day intergalactic magnetic field of sufficient strength. We present a fundamental origin for such a primordial generation of the magnetic field, namely through anomaly cancellation of U1 gauge field in quantum electrodynamics in the context of hilltop inflation. We have analysed at length the power spectrum of the magnetic field, thus generated, which turns out to be helical in nature. We have also found that magnetic power spectrum has significant scaledependence giving rise to a nontrivial magnetic spectral index, a key feature of this model. Interestingly, there exists a large parameter space, where magnetic field of significant strength can be produced. |
Creating a New Persian Poet Based on Machine Learning ; In this article we describe an application of Machine Learning ML and Linguistic Modeling to generate persian poems. In fact we teach machine by reading and learning persian poems to generate fake poems in the same style of the original poems. As two well known poets we used Hafez 13101390 and Saadi 12101292 poems. First we feed the machine with Hafez poems to generate fake poems with the same style and then we feed the machine with the both Hafez and Saadi poems to generate a new style poems which is combination of these two poets styles with emotional Hafez and rational Saadi elements. This idea of combination of different styles with ML opens new gates for extending the treasure of past literature of different cultures. Results show with enough memory, processing power and time it is possible to generate reasonable good poems. |
Mirrorless focusing of XUV highorder harmonics ; By experimentally studying highorder harmonic beams generated in gases, we show how the spatial characteristics of these ultrashort XUV beams can be finely controlled under standard generation conditions. For the first time, we demonstrate that these XUV beams can be emitted as converging beams and get thereby focused after generation. We study this mirrorless focusing using a spatially chirped beam that acts as a spatially localized probe located inside the harmonic generation medium. We analyze the XUV beam evolution with an analytical model providing the beam characteristics and obtain very good agreement with experimental measurements. The XUV foci sizes and positions vary strongly with the harmonic order and the XUV waist can be located at arbitrarily large distances from the generating medium. We discuss how intense XUV fields can be obtained with mirrorless focusing and how such orderdependent XUV beam characteristics are compatible with broadband XUV irradiation and attosecond science. |
Bilinear Adaptive Generalized Vector Approximate Message Passing ; This paper considers the generalized bilinear recovery problem which aims to jointly recover the vector mathbf b and the matrix mathbf X from componentwise nonlinear measurements mathbf Ysim pmathbf Ymathbf Zprodlimitsi,jpYijZij, where mathbf Zmathbf Amathbf bmathbf X, mathbf Acdot is a known affine linear function of mathbf b, and pYijZij is a scalar conditional distribution which models the general output transform. A wide range of realworld applications, e.g., quantized compressed sensing with matrix uncertainty, blind selfcalibration and dictionary learning from nonlinear measurements, onebit matrix completion, etc., can be cast as the generalized bilinear recovery problem. To address this problem, we propose a novel algorithm called the Bilinear Adaptive Generalized Vector Approximate Message Passing BAdGVAMP, which extends the recently proposed Bilinear Adaptive Vector AMP BAdVAMP algorithm to incorporate arbitrary distributions on the output transform. Numerical results on various applications demonstrate the effectiveness of the proposed BAdGVAMP algorithm. |
A Neural Compositional Paradigm for Image Captioning ; Mainstream captioning models often follow a sequential structure to generate captions, leading to issues such as introduction of irrelevant semantics, lack of diversity in the generated captions, and inadequate generalization performance. In this paper, we present an alternative paradigm for image captioning, which factorizes the captioning procedure into two stages 1 extracting an explicit semantic representation from the given image; and 2 constructing the caption based on a recursive compositional procedure in a bottomup manner. Compared to conventional ones, our paradigm better preserves the semantic content through an explicit factorization of semantics and syntax. By using the compositional generation procedure, caption construction follows a recursive structure, which naturally fits the properties of human language. Moreover, the proposed compositional procedure requires less data to train, generalizes better, and yields more diverse captions. |
Learning Gaussian Processes by Minimizing PACBayesian Generalization Bounds ; Gaussian Processes GPs are a generic modelling tool for supervised learning. While they have been successfully applied on large datasets, their use in safetycritical applications is hindered by the lack of good performance guarantees. To this end, we propose a method to learn GPs and their sparse approximations by directly optimizing a PACBayesian bound on their generalization performance, instead of maximizing the marginal likelihood. Besides its theoretical appeal, we find in our evaluation that our learning method is robust and yields significantly better generalization guarantees than other common GP approaches on several regression benchmark datasets. |
User Constrained Thumbnail Generation using Adaptive Convolutions ; Thumbnails are widely used all over the world as a preview for digital images. In this work we propose a deep neural framework to generate thumbnails of any size and aspect ratio, even for unseen values during training, with high accuracy and precision. We use Global Context Aggregation GCA and a modified Region Proposal Network RPN with adaptive convolutions to generate thumbnails in real time. GCA is used to selectively attend and aggregate the global context information from the entire image while the RPN is used to predict candidate bounding boxes for the thumbnail image. Adaptive convolution eliminates the problem of generating thumbnails of various aspect ratios by using filter weights dynamically generated from the aspect ratio information. The experimental results indicate the superior performance of the proposed model over existing stateoftheart techniques. |
Compact Generalized Nonlocal Network ; The nonlocal module is designed for capturing longrange spatiotemporal dependencies in images and videos. Although having shown excellent performance, it lacks the mechanism to model the interactions between positions across channels, which are of vital importance in recognizing finegrained objects and actions. To address this limitation, we generalize the nonlocal module and take the correlations between the positions of any two channels into account. This extension utilizes the compact representation for multiple kernel functions with Taylor expansion that makes the generalized nonlocal module in a fast and lowcomplexity computation flow. Moreover, we implement our generalized nonlocal method within channel groups to ease the optimization. Experimental results illustrate the clearcut improvements and practical applicability of the generalized nonlocal module on both finegrained object recognition and video classification. Code is available at httpsgithub.comKaiyuYuecgnlnetwork.pytorch. |
FARPN Floating Region Proposals for Face Detection ; We propose a novel approach for generating region proposals for performing facedetection. Instead of classifying anchor boxes using features from a pixel in the convolutional feature map, we adopt a poolingbased approach for generating region proposals. However, pooling hundreds of thousands of anchors which are evaluated for generating proposals becomes a computational bottleneck during inference. To this end, an efficient anchor placement strategy for reducing the number of anchorboxes is proposed. We then show that proposals generated by our network Floating Anchor Region Proposal Network, FARPN are better than RPN for generating region proposals for face detection. We discuss several beneficial features of FARPN proposals like iterative refinement, placement of fractional anchors and changing anchors which can be enabled without making any changes to the trained model. Our face detector based on FARPN obtains 89.4 mAP with a ResNet50 backbone on the WIDER dataset. |
Continuous Trajectory Planning Based on Learning Optimization in High Dimensional Input Space for Serial Manipulators ; To continuously generate trajectories for serial manipulators with high dimensional degrees of freedom DOF in the dynamic environment, a realtime optimal trajectory generation method based on machine learning aiming at high dimensional inputs is presented in this paper. First, a learning optimization LO framework is established, and implementations with different submethods are discussed. Additionally, multiple criteria are defined to evaluate the performance of LO models. Furthermore, aiming at high dimensional inputs, a database generation method based on input space dimensionreducing mapping is proposed. At last, this method is validated on motion planning for haptic feedback manipulators HFM in virtual reality systems. Results show that the input space dimensionreducing method can significantly elevate the efficiency and quality of database generation and consequently improve the performance of the LO. Moreover, using this LO method, realtime trajectory generation with high dimensional inputs can be achieved, which lays a foundation for continuous trajectory planning for highDOFrobots in complex environments. |
Thermal optimization of CurzonAhlborn heat engines operating under some generalized efficient power regimes ; In order to establish better performance compromises between the process functionals of a heat engine, in the context of finite time thermodynamics FTT, we propose some generalizations for the well known Efficient Power function through certain variables called Generalization Parameters. These generalization proposals show advantages in the characterization of operation modes for an endoreversible heat engine model. In particular, with introduce the kEfficient Power regime. For this objective function we find the performance of the operation of some power plants through the parameter k. Likewise, for plants that operate in a low efficiency zone, within a configuration space, the k parameter allow us to generate conditions for these plants to operate inside of a high efficiency and low dissipation zone. |
Wav2Pix Speechconditioned Face Generation using Generative Adversarial Networks ; Speech is a rich biometric signal that contains information about the identity, gender and emotional state of the speaker. In this work, we explore its potential to generate face images of a speaker by conditioning a Generative Adversarial Network GAN with raw speech input. We propose a deep neural network that is trained from scratch in an endtoend fashion, generating a face directly from the raw speech waveform without any additional identity information e.g reference image or onehot encoding. Our model is trained in a selfsupervised approach by exploiting the audio and visual signals naturally aligned in videos. With the purpose of training from video data, we present a novel dataset collected for this work, with highquality videos of youtubers with notable expressiveness in both the speech and visual signals. |
Three fermion generations with two unbroken gauge symmetries from the complex sedenions ; We show that three generations of leptons and quarks with unbroken Standard Model gauge symmetry SU3ctimes U1em can be described using the algebra of complexified sedenions mathbbCotimesmathbbS. A primitive idempotent is constructed by selecting a special direction, and the action of this projector on the basis of mathbbCotimesmathbbS can be used to uniquely split the algebra into three complex octonion subalgebras mathbbCotimes mathbbO. These subalgebras all share a common quaternionic subalgebra. The left adjoint actions of the 8 mathbbCdimensional mathbbCotimes mathbbO subalgebras on themselves generates three copies of the Clifford algebra Cell6. It was previously shown that the minimal left ideals of Cell6 describe a single generation of fermions with unbroken SU3ctimes U1em gauge symmetry. Extending this construction from mathbbCotimesmathbbO to mathbbCotimesmathbbS naturally leads to a description of exactly three generations. |
A Generalization Bound for Online Variational Inference ; Bayesian inference provides an attractive onlinelearning framework to analyze sequential data, and offers generalization guarantees which hold even with model mismatch and adversaries. Unfortunately, exact Bayesian inference is rarely feasible in practice and approximation methods are usually employed, but do such methods preserve the generalization properties of Bayesian inference In this paper, we show that this is indeed the case for some variational inference VI algorithms. We consider a few existing online, tempered VI algorithms, as well as a new algorithm, and derive their generalization bounds. Our theoretical result relies on the convexity of the variational objective, but we argue that the result should hold more generally and present empirical evidence in support of this. Our work in this paper presents theoretical justifications in favor of online algorithms relying on approximate Bayesian methods. |
Sizing Storage for Reliable Renewable Integration A Large Deviations Approach ; The inherent intermittency of wind and solar generation presents a significant challenge as we seek to increase the penetration of renewable generation in the power grid. Increasingly, energy storage is being deployed alongside renewable generation to counter this intermittency. However, a formal characterization of the reliability of renewable generators bundled with storage is lacking in the literature. The present paper seeks to fill this gap. We use a Markov modulated fluid queue to model the loss of load probability LOLP associated with a renewable generator bundled with a battery, serving an uncertain demand process. Further, we characterize the asymptotic behavior of the LOLP as the battery size scales to infinity. Our results shed light on the fundamental limits of reliability achievable, and also guide the sizing of the storage required in order to meet a given reliability target. Finally, we present a case study using realworld wind power data to demonstrate the applicability of our results in practice. |
Generalized Supergravity Equations and Generalized FradkinTseytlin Counterterm ; The generalized FradkinTseytlin counterterm for the type I GreenSchwarz superstring is determined for background fields satisfying the generalized supergravity equations GSE. For this purpose, we revisit the derivation of the GSE based upon the requirement of kappasymmetry of the superstring action. Lifting the constraint of vanishing bosonic torsion components, we are able to make contact to several different torsion constraints used in the literature. It is argued that a natural geometric interpretation of the GSE vector field that generalizes the dilaton is as the torsion vector, which can combine with the dilatino spinor into the torsion supervector. To find the counterterm, we use old results for the oneloop effective action of the heterotic sigma model. The counterterm is covariant and involves the worldsheet torsion for vanishing curvature, but cannot be constructed as a local functional in terms of the worldsheet metric. It is shown that the Weyl anomaly cancels without imposing any further constraints on the background fields. In the case of ordinary supergravity, it reduces to the FradkinTseytlin counterterm modulo an additional constraint. |
Distribution System State Estimation in the Presence of High Solar Penetration ; Lowtomedium voltage distribution networks are experiencing rising levels of distributed energy resources, including renewable generation, along with improved sensing, communication, and automation infrastructure. As such, state estimation methods for distribution systems are becoming increasingly relevant as a means to enable better control strategies that can both leverage the benefits and mitigate the risks associated with high penetration of variable and uncertain distributed generation resources. The primary challenges of this problem include modeling complexities nonlinear, nonconvex powerflow equations, limited availability of sensor measurements, and high penetration of uncertain renewable generation. This paper formulates the distribution system state estimation as a nonlinear, weighted, least squares problem, based on sensor measurements as well as forecast data both load and generation. We investigate the sensitivity of state estimator accuracy to loadgeneration forecast uncertainties, sensor accuracy, and sensor coverage levels. |
Multispecies Stochastic Model And Effective Stochastic Generator with SiteDependent Interactions ; The dynamical rules in auxiliary stochastic process that generates the biased ensemble of rare events are nonlocal. For the systems with one type of particle, it is shown that there are special cases for which the generators of effective processes can include local interactions. In this paper we investigate this possibility for a systems of classical particles with more than one type of particle moving on a onedimensional lattice with open boundary conditions. Assuming that the interactions in the original process are local and siteindependent and also it is assumed that the particles have hardcore interactions. We will show that under certain constraints on the microscopic reaction rules, the stochastic generator of unconditioned process can be local but sitedependent. As one examples, Amodel with two species of particles are presented and be investigated the constraints under which the effective generators are local and sitedependent. |
Attending to Future Tokens For Bidirectional Sequence Generation ; Neural sequence generation is typically performed tokenbytoken and lefttoright. Whenever a token is generated only previously produced tokens are taken into consideration. In contrast, for problems such as sequence classification, bidirectional attention, which takes both past and future tokens into consideration, has been shown to perform much better. We propose to make the sequence generation process bidirectional by employing special placeholder tokens. Treated as a node in a fully connected graph, a placeholder token can take past and future tokens into consideration when generating the actual output token. We verify the effectiveness of our approach experimentally on two conversational tasks where the proposed bidirectional model outperforms competitive baselines by a large margin. |
CIMBA fast Monte Carlo generation using cubic interpolation ; Monte Carlo generation of high energy particle collisions is a critical tool for both theoretical and experimental particle physics, connecting perturbative calculations to phenomenological models, and theory predictions to full detector simulation. The generation of minimum bias events can be particularly computationally expensive, where nonperturbative effects play an important role and specific processes and fiducial regions can no longer be well defined. In particular scenarios, particle guns can be used to quickly sample kinematics for single particles produced in minimum bias events. CIMBA Cubic Interpolation for Minimum Bias Approximation provides a comprehensive package to smoothly sample predefined kinematic grids, from any general purpose Monte Carlo generator, for all particles produced in minimum bias events. These grids are provided for a number of beam configurations including those of the Large Hadron Collider. |
The Limitations of Stylometry for Detecting MachineGenerated Fake News ; Recent developments in neural language models LMs have raised concerns about their potential misuse for automatically spreading misinformation. In light of these concerns, several studies have proposed to detect machinegenerated fake news by capturing their stylistic differences from humanwritten text. These approaches, broadly termed stylometry, have found success in source attribution and misinformation detection in humanwritten texts. However, in this work, we show that stylometry is limited against machinegenerated misinformation. While humans speak differently when trying to deceive, LMs generate stylistically consistent text, regardless of underlying motive. Thus, though stylometry can successfully prevent impersonation by identifying text provenance, it fails to distinguish legitimate LM applications from those that introduce false information. We create two benchmarks demonstrating the stylistic similarity between malicious and legitimate uses of LMs, employed in autocompletion and editingassistance settings. Our findings highlight the need for nonstylometry approaches in detecting machinegenerated misinformation, and open up the discussion on the desired evaluation benchmarks. |
Leveraging Frequent Query Substructures to Generate Formal Queries for Complex Question Answering ; Formal query generation aims to generate correct executable queries for question answering over knowledge bases KBs, given entity and relation linking results. Current approaches build universal paraphrasing or ranking models for the whole questions, which are likely to fail in generating queries for complex, longtail questions. In this paper, we propose SubQG, a new query generation approach based on frequent query substructures, which helps rank the existing but nonsignificant query structures or build new query structures. Our experiments on two benchmark datasets show that our approach significantly outperforms the existing ones, especially for complex questions. Also, it achieves promising performance with limited training data and noisy entityrelation linking results. |
Generating Personalized Recipes from Historical User Preferences ; Existing approaches to recipe generation are unable to create recipes for users with culinary preferences but incomplete knowledge of ingredients in specific dishes. We propose a new task of personalized recipe generation to help these users expanding a name and incomplete ingredient details into complete naturaltext instructions aligned with the user's historical preferences. We attend on technique and recipelevel representations of a user's previously consumed recipes, fusing these 'useraware' representations in an attention fusion layer to control recipe text generation. Experiments on a new dataset of 180K recipes and 700K interactions show our model's ability to generate plausible and personalized recipes compared to nonpersonalized baselines. |
Dynamic State Estimation of Generators Under Cyber Attacks ; Accurate and reliable estimation of generator's dynamic state vectors in real time are critical to the monitoring and control of power systems. A robust Cubature Kalman Filter RCKF based approach is proposed for dynamic state estimation DSE of generators under cyber attacks in this paper. First, two types of cyber attacks, namely false data injection and denial of service attacks, are modelled and thereby introduced into DSE of a generator by mixing the attack vectors with the measurement data; Second, under cyber attacks with different degrees of sophistication, the RCKF algorithm and the Cubature Kalman Filter CKF algorithm are adopted to the DSE, and then the two algorithms are compared and discussed. The novelty of this study lies primarily in our attempt to introduce cyber attacks into DSE of generators. The simulation results on the IEEE 9bus system and the New England 16machine 68bus system verify the effectiveness and superiority of the RCKF. |
Improving Generalization by Incorporating Coverage in Natural Language Inference ; The task of natural language inference NLI is to identify the relation between the given premise and hypothesis. While recent NLI models achieve very high performance on individual datasets, they fail to generalize across similar datasets. This indicates that they are solving NLI datasets instead of the task itself. In order to improve generalization, we propose to extend the input representations with an abstract view of the relation between the hypothesis and the premise, i.e., how well the individual words, or word ngrams, of the hypothesis are covered by the premise. Our experiments show that the use of this information considerably improves generalization across different NLI datasets without requiring any external knowledge or additional data. Finally, we show that using the coverage information is not only beneficial for improving the performance across different datasets of the same task. The resulting generalization improves the performance across datasets that belong to similar but not the same tasks. |
Constantroll inflation in scalartensor gravity ; We generalize the notion of constantroll inflation earlier introduced in General Relativity GR and fR gravity to inflationary models in more general scalartensor gravity. A number of novel exact analytic solutions for a FLRW spatially flat cosmological background is found for this case. All forms of the scalar field potential and its coupling to gravity producing the exact de Sitter solution, while the scalar field is varying, are presented. In the particular cases of induced gravity and GR with a nonminimally coupled scalar field, all constantroll inflationary solutions are found. In the former case they represent powerlaw inflation, while in the latter case the solution is novel and more complicated. Comparison of scalar perturbations generated during such inflation in induced gravity with observational data shows that the constantroll parameter should be small, similar to constantroll inflation in GR and fR gravity. Then the solution reduces to the standard slowroll one with small corrections. |
FullScale Continuous Synthetic Sonar Data Generation with Markov Conditional Generative Adversarial Networks ; Deployment and operation of autonomous underwater vehicles is expensive and timeconsuming. Highquality realistic sonar data simulation could be of benefit to multiple applications, including training of human operators for postmission analysis, as well as tuning and validation of autonomous target recognition ATR systems for underwater vehicles. Producing realistic synthetic sonar imagery is a challenging problem as the model has to account for specific artefacts of real acoustic sensors, vehicle altitude, and a variety of environmental factors. We propose a novel method for generating realisticlooking sonar sidescans of fulllength missions, called Markov Conditional pix2pix MCpix2pix. Quantitative assessment results confirm that the quality of the produced data is almost indistinguishable from real. Furthermore, we show that bootstrapping ATR systems with MCpix2pix data can improve the performance. Synthetic data is generated 18 times faster than real acquisition speed, with full user control over the topography of the generated data. |
Transform the Set Memory Attentive Generation of Guided and Unguided Image Collages ; Cutting and pasting image segments feels intuitive the choice of source templates gives artists flexibility in recombining existing source material. Formally, this process takes an image set as input and outputs a collage of the set elements. Such selection from sets of source templates does not fit easily in classical convolutional neural models requiring inputs of fixed size. Inspired by advances in attention and setinput machine learning, we present a novel architecture that can generate in one forward pass image collages of source templates using setstructured representations. This paper has the following contributions i a novel framework for image generation called Memory Attentive Generation of Image Collages MAGIC which gives artists new ways to create digital collages; ii from the machinelearning perspective, we show a novel Generative Adversarial Networks GAN architecture that uses SetTransformer layers and setpooling to blend sets of random image samples a hybrid nonparametric approach. |
DwNet Dense warpbased network for poseguided human video generation ; Generation of realistic highresolution videos of human subjects is a challenging and important task in computer vision. In this paper, we focus on human motion transfer generation of a video depicting a particular subject, observed in a single image, performing a series of motions exemplified by an auxiliary driving video. Our GANbased architecture, DwNet, leverages dense intermediate poseguided representation and refinement process to warp the required subject appearance, in the form of the texture, from a source image into a desired pose. Temporal consistency is maintained by further conditioning the decoding process within a GAN on the previously generated frame. In this way a video is generated in an iterative and recurrent fashion. We illustrate the efficacy of our approach by showing stateoftheart quantitative and qualitative performance on two benchmark datasets TaiChi and Fashion Modeling. The latter is collected by us and will be made publicly available to the community. |
Bridging the Gap Between fGANs and Wasserstein GANs ; Generative adversarial networks GANs have enjoyed much success in learning highdimensional distributions. Learning objectives approximately minimize an fdivergence fGANs or an integral probability metric Wasserstein GANs between the model and the data distribution using a discriminator. Wasserstein GANs enjoy superior empirical performance, but in fGANs the discriminator can be interpreted as a density ratio estimator which is necessary in some GAN applications. In this paper, we bridge the gap between fGANs and Wasserstein GANs WGANs. First, we list two constraints over variational fdivergence estimation objectives that preserves the optimal solution. Next, we minimize over a Lagrangian relaxation of the constrained objective, and show that it generalizes critic objectives of both fGAN and WGAN. Based on this generalization, we propose a novel practical objective, named KLWasserstein GAN KLWGAN. We demonstrate empirical success of KLWGAN on synthetic datasets and realworld image generation benchmarks, and achieve stateoftheart FID scores on CIFAR10 image generation. |
Retrieve and Refine Exemplarbased Neural Comment Generation ; Code comment generation is a crucial task in the field of automatic software development. Most previous neural comment generation systems used an encoderdecoder neural network and encoded only information from source code as input. Software reuse is common in software development. However, this feature has not been introduced to existing systems. Inspired by the traditional IRbased approaches, we propose to use the existing comments of similar source code as exemplars to guide the comment generation process. Based on an open source search engine, we first retrieve a similar code and treat its comment as an exemplar. Then we applied a seq2seq neural network to conduct an exemplarbased comment generation. We evaluate our approach on a largescale Java corpus, and experimental results demonstrate that our model significantly outperforms the stateoftheart methods. |
Secret Key Generation via PulseCoupled Synchronization ; A novel framework for sharing common randomness and generating secret keys in wireless networks is considered. In particular, a network of users equipped with pulse oscillators POs and coupling mechanisms in between is considered. Such mechanisms exist in synchronized biological and natural systems, and have been exploited to provide synchronization in distributed networks. We show that naturallyexisting initial random phase differences between the POs in the network can be utilized to provide almost identical common randomness to the users. This randomness is extracted from the synchronization time in the network. Bounds on the entropy of such randomness are derived for a twouser system and a conjecture is made for a general nuser system. Then, a threeterminal scenario is considered including two legitimate users and a passive eavesdropper, referred to as Eve. Since in a practical setting Eve receives pulses with propagation delays, she can not identify the exact synchronization time. A simplified model is then considered for Eve's receiver and then a bound on the rate of secret key generation is derived. Also, it is shown, under certain conditions, that the proposed protocol is resilient to an active jammer equipped with a similar pulse generation mechanism. |
Transferring neural speech waveform synthesizers to musical instrument sounds generation ; Recent neural waveform synthesizers such as WaveNet, WaveGlow, and the neuralsourcefilter NSF model have shown good performance in speech synthesis despite their different methods of waveform generation. The similarity between speech and music audio synthesis techniques suggests interesting avenues to explore in terms of the best way to apply speech synthesizers in the music domain. This work compares three neural synthesizers used for musical instrument sounds generation under three scenarios training from scratch on music data, zeroshot learning from the speech domain, and finetuningbased adaptation from the speech to the music domain. The results of a largescale perceptual test demonstrated that the performance of three synthesizers improved when they were pretrained on speech data and finetuned on music data, which indicates the usefulness of knowledge from speech data for music audio generation. Among the synthesizers, WaveGlow showed the best potential in zeroshot learning while NSF performed best in the other scenarios and could generate samples that were perceptually close to natural audio. |
SketchFillAR A PersonaGrounded ChitChat Generation Framework ; Humanlike chitchat conversation requires agents to generate responses that are fluent, engaging and consistent. We propose SketchFillAR, a framework that uses a personamemory to generate chitchat responses in three phases. First, it generates dynamic sketch responses with open slots. Second, it generates candidate responses by filling slots with parts of its stored persona traits. Lastly, it ranks and selects the final response via a language model score. SketchFillAR outperforms a stateoftheart baseline both quantitatively 10point lower perplexity and qualitatively preferred by 55 headsup in singleturn and 20 higher in consistency in multiturn user studies on the PersonaChat dataset. Finally, we extensively analyze SketchFillAR's responses and human feedback, and show it is more consistent and engaging by using more relevant responses and questions. |
Incorporating InterlocutorAware Context into Response Generation on MultiParty Chatbots ; Conventional chatbots focus on twoparty response generation, which simplifies the real dialogue scene. In this paper, we strive toward a novel task of Response Generation on MultiParty Chatbot RGMPC, where the generated responses heavily rely on the interlocutors' roles e.g., speaker and addressee and their utterances. Unfortunately, complex interactions among the interlocutors' roles make it challenging to precisely capture conversational contexts and interlocutors' information. Facing this challenge, we present a response generation model which incorporates Interlocutoraware Contexts into Recurrent EncoderDecoder frameworks ICRED for RGMPC. Specifically, we employ interactive representations to capture dialogue contexts for different interlocutors. Moreover, we leverage an addressee memory to enhance contextual interlocutor information for the target addressee. Finally, we construct a corpus for RGMPC based on an existing openaccess dataset. Automatic and manual evaluations demonstrate that the ICRED remarkably outperforms strong baselines. |
Asymptotically unbiased estimation of physical observables with neural samplers ; We propose a general framework for the estimation of observables with generative neural samplers focusing on modern deep generative neural networks that provide an exact sampling probability. In this framework, we present asymptotically unbiased estimators for generic observables, including those that explicitly depend on the partition function such as free energy or entropy, and derive corresponding variance estimators. We demonstrate their practical applicability by numerical experiments for the 2d Ising model which highlight the superiority over existing methods. Our approach greatly enhances the applicability of generative neural samplers to realworld physical systems. |
Study of Constrained Network Structures for WGANs on Numeric Data Generation ; Some recent studies have suggested using GANs for numeric data generation such as to generate data for completing the imbalanced numeric data. Considering the significant difference between the dimensions of the numeric data and images, as well as the strong correlations between features of numeric data, the conventional GANs normally face an overfitting problem, consequently leads to an illconditioning problem in generating numeric and structured data. This paper studies the constrained network structures between generator G and discriminator D in WGAN, designs several structures including isomorphic, mirror and selfsymmetric structures. We evaluates the performances of the constrained WGANs in data augmentations, taking the nonconstrained GANs and WGANs as the baselines. Experiments prove the constrained structures have been improved in 1720 groups of experiments. In twenty experiments on four UCI Machine Learning Repository datasets, Australian Credit Approval data, German Credit data, Pima Indians Diabetes data and SPECT heart data facing five conventional classifiers. Especially, Isomorphic WGAN is the best in 1520 experiments. Finally, we theoretically proves that the effectiveness of constrained structures by the directed graphic model DGM analysis. |
GANenhanced Conditional Echocardiogram Generation ; Echocardiography echo is a common means of evaluating cardiac conditions. Due to the label scarcity, semisupervised paradigms in automated echo analysis are getting traction. One of the most soughtafter problems in echo is the segmentation of cardiac structures e.g. chambers. Accordingly, we propose an echocardiogram generation approach using generative adversarial networks with a conditional patchbased discriminator. In this work, we validate the feasibility of GANenhanced echo generation with different conditions segmentation masks, namely, the left ventricle, ventricular myocardium, and atrium. Results show that the proposed adversarial algorithm can generate highquality echo frames whose cardiac structures match the given segmentation masks. This method is expected to facilitate the training of other machine learning models in a semisupervised fashion as suggested in similar researches. |
EmpDG Multiresolution Interactive Empathetic Dialogue Generation ; A humanized dialogue system is expected to generate empathetic replies, which should be sensitive to the users' expressed emotion. The task of empathetic dialogue generation is proposed to address this problem. The essential challenges lie in accurately capturing the nuances of human emotion and considering the potential of user feedback, which are overlooked by the majority of existing work. In response to this problem, we propose a multiresolution adversarial model EmpDG, to generate more empathetic responses. EmpDG exploits both the coarsegrained dialoguelevel and finegrained tokenlevel emotions, the latter of which helps to better capture the nuances of user emotion. In addition, we introduce an interactive adversarial learning framework which exploits the user feedback, to identify whether the generated responses evoke emotion perceptivity in dialogues. Experimental results show that the proposed approach significantly outperforms the stateoftheart baselines in both content quality and emotion perceptivity. |
Helicity amplitudes for generic multibody particle decays featuring multiple decay chains ; We present the general expression of helicity amplitudes for generic multibody particle decays characterised by multiple decay chains. This is achieved by addressing for the first time the issue of the matching of final particle spin states among different decay chains in full generality for generic multibody decays, proposing a method able to match the exact definition of spin states relative to the decaying particle ones. We stress the importance of our result by showing that one of the matching method used in the literature is incorrect, leading to amplitude models violating rotational invariance. The results presented are therefore relevant for performing numerous amplitude analysis, notably those searching for exotic structures like pentaquarks. |
Aggregative Efficiency of Bayesian Learning in Networks ; When individuals in a social network learn about an unknown state from private signals and neighbors' actions, the network structure often causes information loss. We consider rational agents and Gaussian signals in the canonical sequential sociallearning problem and ask how the network changes the efficiency of signal aggregation. Rational actions in our model are loglinear functions of observations and admit a signalcounting interpretation of accuracy. Networks where agents observe multiple neighbors but not their common predecessors confound information, and even a small amount of confounding can lead to much lower accuracy. In a class of networks where agents move in generations and observe the previous generation, we quantify the information loss with an aggregative efficiency index. Aggregative efficiency is a simple function of network parameters increasing in observations and decreasing in confounding. Later generations contribute little additional information, even with arbitrarily large generations. |
GANkyoku a Generative Adversarial Network for Shakuhachi Music ; A common approach to generating symbolic music using neural networks involves repeated sampling of an autoregressive model until the full output sequence is obtained. While such approaches have shown some promise in generating short sequences of music, this typically has not extended to cases where the final target sequence is significantly longer, for example an entire piece of music. In this work we propose a network trained in an adversarial process to generate entire pieces of solo shakuhachi music, in the form of symbolic notation. The pieces are intended to refer clearly to traditional shakuhachi music, maintaining idiomaticity and key aesthetic qualities, while also adding novel features, ultimately creating worthy additions to the contemporary shakuhachi repertoire. A key subproblem is also addressed, namely the lack of relevant training data readily available, in two steps firstly, we introduce the PHShaku dataset for symbolic traditional shakuhachi music; secondly, we build on previous work using conditioning in generative adversarial networks to introduce a technique for data augmentation. |
KPTimes A LargeScale Dataset for Keyphrase Generation on News Documents ; Keyphrase generation is the task of predicting a set of lexical units that conveys the main content of a source text. Existing datasets for keyphrase generation are only readily available for the scholarly domain and include nonexpert annotations. In this paper we present KPTimes, a largescale dataset of news texts paired with editorcurated keyphrases. Exploring the dataset, we show how editors tag documents, and how their annotations differ from those found in existing datasets. We also train and evaluate stateoftheart neural keyphrase generation models on KPTimes to gain insights on how well they perform on the news domain. The dataset is available online at httpsgithub.comygorgKPTimes . |
ClassConditional VAEGAN for LocalAncestry Simulation ; Local ancestry inference LAI allows identification of the ancestry of all chromosomal segments in admixed individuals, and it is a critical step in the analysis of human genomes with applications from pharmacogenomics and precision medicine to genomewide association studies. In recent years, many LAI techniques have been developed in both industry and academic research. However, these methods require large training data sets of human genomic sequences from the ancestries of interest. Such reference data sets are usually limited, proprietary, protected by privacy restrictions, or otherwise not accessible to the public. Techniques to generate training samples that resemble real haploid sequences from ancestries of interest can be useful tools in such scenarios, since a generalized model can often be shared, but the unique human sample sequences cannot. In this work we present a classconditional VAEGAN to generate new human genomic sequences that can be used to train local ancestry inference LAI algorithms. We evaluate the quality of our generated data by comparing the performance of a stateoftheart LAI method when trained with generated versus real data. |
DEGAS Differentiable Efficient Generator Search ; Network architecture search NAS achieves stateoftheart results in various tasks such as classification and semantic segmentation. Recently, a reinforcement learningbased approach has been proposed for Generative Adversarial Networks GANs search. In this work, we propose an alternative strategy for GAN search by using a method called DEGAS Differentiable Efficient GenerAtor Search, which focuses on efficiently finding the generator in the GAN. Our search algorithm is inspired by the differential architecture search strategy and the Global Latent Optimization GLO procedure. This leads to both an efficient and stable GAN search. After the generator architecture is found, it can be plugged into any existing framework for GAN training. For CTGAN, which we use in this work, the new model outperforms the original inception score results by 0.25 for CIFAR10 and 0.77 for STL. It also gets better results than the RL based GAN search methods in shorter search time. |
The DCTC condition is generically fulfilled in classical nonquantum statistical systems ; The DCTC condition, introduced by David Deutsch as a condition to be fulfilled by analogues for processes of quantum systems in the presence of closed timelike curves, is investigated for classical statistical nonquantum bipartite systems. It is shown that the DCTC condition can generically be fulfilled in classical statistical systems, under very general, modelindependent conditions. The central property used is the convexity and completeness of the state space that allows it to generalize Deutsch's original proof for qbit systems to more general classes of statistically described systems. The results demonstrate that the DCTC condition, or the conditions under which it can be fulfilled, is not characteristic of, or dependent on, the quantum nature of a bipartite system. |
Quantum Maximin Surfaces ; We formulate a quantum generalization of maximin surfaces and show that a quantum maximin surface is identical to the minimal quantum extremal surface, introduced in the EW prescription. We discuss various subtleties and complications associated to a maximinimization of the bulk von Neumann entropy due to corners and unboundedness and present arguments that nonetheless a maximinimization of the UVfinite generalized entropy should be welldefined. We give the first general proof that the EW prescription satisfies entanglement wedge nesting and the strong subadditivity inequality. In addition, we apply the quantum maximin technology to prove that recently proposed generalizations of the EW prescription to nonholographic subsystems including the socalled quantum extremal islands also satisfy entanglement wedge nesting and strong subadditivity. Our results hold in the regime where backreaction of bulk quantum fields can be treated perturbatively in GNhbar, but we emphasize that they are valid even when gradients of the bulk entropy are of the same order as variations in the area, a regime recently investigated in new models of black hole evaporation in AdSCFT. |
Generalizing the SokolovTernov effect for radiative polarization in intense laser fields ; A consistent description of the radiative polarization for relativistic electrons in intense laser fields is derived by generalizing the SokolovTernov effect in general field structure. The new form together with the spindependent radiationreaction force provides a complete set of dynamical equations for electron momentum and spin in strong fields. When applied to varying intense fields, e.g. the laser fields, the generalized SokolovTernov effect allows electrons to gain or lose polarization in any directions other than along the magnetic field in the rest frame of the electron. The generalized theory is applied to the collision process between initially polarizedunpolarized high energy electrons with linearly polarized ultraintense laser pulse, showing results that eliminate the dependence on specific choices of a quantization axis and spin initialization existing in spinprojection models. |
Bayesian highdimensional linear regression with generic spikeandslab priors ; Spikeandslab priors are popular Bayesian solutions for highdimensional linear regression problems. Previous theoretical studies on spikeandslab methods focus on specific prior formulations and use priordependent conditions and analyses, and thus can not be generalized directly. In this paper, we propose a class of generic spikeandslab priors and develop a unified framework to rigorously assess their theoretical properties. Technically, we provide general conditions under which generic spikeandslab priors can achieve the nearlyoptimal posterior contraction rate and the model selection consistency. Our results include those of Narisetty and He 2014 and Castillo et al. 2015 as special cases. |
MM for Penalized Estimation ; Penalized estimation can conduct variable selection and parameter estimation simultaneously. The general framework is to minimize a loss function subject to a penalty designed to generate sparse variable selection. The majorizationminimization MM algorithm is a computational scheme for stability and simplicity, and the MM algorithm has been widely applied in penalized estimation. Much of the previous work have focused on convex loss functions such as generalized linear models. When data are contaminated with outliers, robust loss functions can generate more reliable estimates. Recent literature has witnessed a growing impact of nonconvex lossbased methods, which can generate robust estimation for data contaminated with outliers. This article investigates MM algorithm for penalized estimation, provide innovative optimality conditions and establish convergence theory with both convex and nonconvex loss functions. With respect to applications, we focus on several nonconvex loss functions, which were formerly studied in machine learning for regression and classification problems. Performance of the proposed algorithms are evaluated on simulated and real data including healthcare costs and cancer clinical status. Efficient implementations of the algorithms are available in the R package mpath in CRAN. |
Generating Object Stamps ; We present an algorithm to generate diverse foreground objects and composite them into background images using a GAN architecture. Given an object class, a userprovided bounding box, and a background image, we first use a mask generator to create an object shape, and then use a texture generator to fill the mask such that the texture integrates with the background. By separating the problem of object insertion into these two stages, we show that our model allows us to improve the realism of diverse object generation that also agrees with the provided background image. Our results on the challenging COCO dataset show improved overall quality and diversity compared to stateoftheart object insertion approaches. |
Generative Pseudolabel Refinement for Unsupervised Domain Adaptation ; We investigate and characterize the inherent resilience of conditional Generative Adversarial Networks cGANs against noise in their conditioning labels, and exploit this fact in the context of Unsupervised Domain Adaptation UDA. In UDA, a classifier trained on the labelled source set can be used to infer pseudolabels on the unlabelled target set. However, this will result in a significant amount of misclassified examples due to the wellknown domain shift issue, which can be interpreted as noise injection in the groundtruth labels for the target set. We show that cGANs are, to some extent, robust against such shift noise. Indeed, cGANs trained with noisy pseudolabels, are able to filter such noise and generate cleaner target samples. We exploit this finding in an iterative procedure where a generative model and a classifier are jointly trained in turn, the generator allows to sample cleaner data from the target distribution, and the classifier allows to associate better labels to target samples, progressively refining target pseudolabels. Results on common benchmarks show that our method performs better or comparably with the unsupervised domain adaptation state of the art. |
Mass generation by fractional instantons in SUn chains ; Recently, Haldane's conjecture about SU2 chains has been generalized to SUn chains in the symmetric representations. For a rankp representation, a gapless phase is predicted when p and n are coprime; otherwise, a finite energy gap is present above the ground state. In this work, we offer an intuitive explanation of this behavior based on fractional topological excitations, which are able to generate a mass gap except when p and n have no common divisor. This is a generalization of an older work in SU2, which explains the generation of the Haldane gap in terms of merons in the O3 nonlinear sigma model. |
Generating Natural Adversarial Hyperspectral examples with a modified Wasserstein GAN ; Adversarial examples are a hot topic due to their abilities to fool a classifier's prediction. There are two strategies to create such examples, one uses the attacked classifier's gradients, while the other only requires access to the classifier's prediction. This is particularly appealing when the classifier is not full known black box model. In this paper, we present a new method which is able to generate natural adversarial examples from the true data following the second paradigm. Based on Generative Adversarial Networks GANs 5, it reweights the true data empirical distribution to encourage the classifier to generate adversarial examples. We provide a proof of concept of our method by generating adversarial hyperspectral signatures on a remote sensing dataset. |
Diagrammatic Monte Carlo Method for Impurity Models with General Interactions and Hybridizations ; We present a diagrammatic Monte Carlo method for quantum impurity problems with general interactions and general hybridization functions. Our method uses a recursive determinant scheme to sample diagrams for the scattering amplitude. Unlike in other methods for general impurity problems, an approximation of the continuous hybridization function by a finite number of bath states is not needed, and accessing low temperature does not incur an exponential cost. We test the method for the example of molecular systems, where we systematically vary temperature, interatomic distance, and basis set size. We further apply the method to an impurity problem generated by a selfenergy embedding calculation of correlated antiferromagnetic NiO. We find that the method is ideal for quantum impurity problems with a large number of orbitals but only moderate correlations. |
Machine Translation Pretraining for DatatoText Generation A Case Study in Czech ; While there is a large body of research studying deep learning methods for text generation from structured data, almost all of it focuses purely on English. In this paper, we study the effectiveness of machine translation based pretraining for datatotext generation in nonEnglish languages. Since the structured data is generally expressed in English, text generation into other languages involves elements of translation, transliteration and copying elements already encoded in neural machine translation systems. Moreover, since datatotext corpora are typically small, this task can benefit greatly from pretraining. Based on our experiments on Czech, a morphologically complex language, we find that pretraining lets us train endtoend models with significantly improved performance, as judged by automatic metrics and human evaluation. We also show that this approach enjoys several desirable properties, including improved performance in low data scenarios and robustness to unseen slot values. |
Predicting Camera Viewpoint Improves Crossdataset Generalization for 3D Human Pose Estimation ; Monocular estimation of 3d human pose has attracted increased attention with the availability of large groundtruth motion capture datasets. However, the diversity of training data available is limited and it is not clear to what extent methods generalize outside the specific datasets they are trained on. In this work we carry out a systematic study of the diversity and biases present in specific datasets and its effect on crossdataset generalization across a compendium of 5 pose datasets. We specifically focus on systematic differences in the distribution of camera viewpoints relative to a bodycentered coordinate frame. Based on this observation, we propose an auxiliary task of predicting the camera viewpoint in addition to pose. We find that models trained to jointly predict viewpoint and pose systematically show significantly improved crossdataset generalization. |
The general theory of permutation equivarant neural networks and higher order graph variational encoders ; Previous work on symmetric group equivariant neural networks generally only considered the case where the group acts by permuting the elements of a single vector. In this paper we derive formulae for general permutation equivariant layers, including the case where the layer acts on matrices by permuting their rows and columns simultaneously. This case arises naturally in graph learning and relation learning applications. As a specific case of higher order permutation equivariant networks, we present a second order graph variational encoder, and show that the latent distribution of equivariant generative models must be exchangeable. We demonstrate the efficacy of this architecture on the tasks of link prediction in citation graphs and molecular graph generation. |
Paralleldistributed implementation of cellular training for generative adversarial neural networks ; Generative adversarial networks GANs are widely used to learn generative models. GANs consist of two networks, a generator and a discriminator, that apply adversarial learning to optimize their parameters. This article presents a paralleldistributed implementation of a cellular competitive coevolutionary method to train two populations of GANs. A distributed memory parallel implementation is proposed for execution in high performancesupercomputing centers. Efficient results are reported on addressing the generation of handwritten digits MNIST dataset samples. Moreover, the proposed implementation is able to reduce the training times and scale properly when considering different grid sizes for training. |
Generalized Multivariate Hawkes Processes ; This work contributes to the theory and applications of Hawkes processes. We introduce and examine a new class of Hawkes processes that we call generalized Hawkes processes, and their special subclass the generalized multivariate Hawkes processes GMHPs. GMHPs are multivariate marked point processes that add an important feature to the family of the classical multivariate Hawkes processes they allow for explicit modelling of simultaneous occurrence of excitation events coming from different sources, i.e. caused by different coordinates of the multivariate process. We study the issue of existence of a generalized Hawkes process, and we provide a construction of a specific generalized multivariate Hawkes process. We investigate Markovian aspects of GMHPs, and we indicate some plausible important applications of GMHPs. |
Stay Hungry, Stay Focused Generating Informative and Specific Questions in InformationSeeking Conversations ; We investigate the problem of generating informative questions in informationasymmetric conversations. Unlike previous work on question generation which largely assumes knowledge of what the answer might be, we are interested in the scenario where the questioner is not given the context from which answers are drawn, but must reason pragmatically about how to acquire new information, given the shared conversation history. We identify two core challenges 1 formally defining the informativeness of potential questions, and 2 exploring the prohibitively large space of potential questions to find the good candidates. To generate pragmatic questions, we use reinforcement learning to optimize an informativeness metric we propose, combined with a reward function designed to promote more specific questions. We demonstrate that the resulting pragmatic questioner substantially improves the informativeness and specificity of questions generated over a baseline model, as evaluated by our metrics as well as humans. |
Towards Unsupervised Language Understanding and Generation by Joint Dual Learning ; In modular dialogue systems, natural language understanding NLU and natural language generation NLG are two critical components, where NLU extracts the semantics from the given texts and NLG is to construct corresponding natural language sentences based on the input semantic representations. However, the dual property between understanding and generation has been rarely explored. The prior work is the first attempt that utilized the duality between NLU and NLG to improve the performance via a dual supervised learning framework. However, the prior work still learned both components in a supervised manner, instead, this paper introduces a general learning framework to effectively exploit such duality, providing flexibility of incorporating both supervised and unsupervised learning algorithms to train language understanding and generation models in a joint fashion. The benchmark experiments demonstrate that the proposed approach is capable of boosting the performance of both NLU and NLG. |
Extensions of Dupire Formula Stochastic Interest Rates and Stochastic Local Volatility ; We derive generalizations of Dupire formula to the cases of general stochastic drift andor stochastic local volatility. First, we handle a case in which the drift is given as difference of two stochastic short rates. Such a setting is natural in foreign exchange context where the short rates correspond to the short rates of the two currencies, equity singlecurrency context with stochastic dividend yield, or commodity context with stochastic convenience yield. We present the formula both in a call surface formulation as well as total implied variance formulation where the latter avoids calendar spread arbitrage by construction. We provide derivations for the case where both short rates are given as single factor processes and present the limits for a single stochastic rate or all deterministic short rates. The limits agree with published results. Then we derive a formulation that allows a more general stochastic drift and diffusion including one or more stochastic local volatility terms. In the general setting, our derivation allows the computation and calibration of the leverage function for stochastic local volatility models. Despite being implicit, the generalized Dupire formulae can be used numerically in a fixedpoint iterative scheme. |
Integral representation for energies in linear elasticity with surface discontinuities ; In this paper we prove an integral representation formula for a general class of energies defined on the space of generalized special functions of bounded deformation GSBDp in arbitrary space dimensions. Functionals of this type naturally arise in the modeling of linear elastic solids with surface discontinuities including phenomena as fracture, damage, surface tension between different elastic phases, or material voids. Our approach is based on the global method for relaxation devised in Bouchitte et al. '98 and a recent Korntype inequality in GSBDp CagnettiChambolleScardia '20. Our general strategy also allows to generalize integral representation results in SBDp, obtained in dimension two ContiFocardiIurlano '16, to higher dimensions, and to revisit results in the framework of generalized special functions of bounded variation GSBVp. |
BeCAPTCHA Behavioral Bot Detection using Touchscreen and Mobile Sensors benchmarked on HuMIdb ; In this paper we study the suitability of a new generation of CAPTCHA methods based on smartphone interactions. The heterogeneous flow of data generated during the interaction with the smartphones can be used to model human behavior when interacting with the technology and improve bot detection algorithms. For this, we propose BeCAPTCHA, a CAPTCHA method based on the analysis of the touchscreen information obtained during a single drag and drop task in combination with the accelerometer data. The goal of BeCAPTCHA is to determine whether the drag and drop task was realized by a human or a bot. We evaluate the method by generating fake samples synthesized with Generative Adversarial Neural Networks and handcrafted methods. Our results suggest the potential of mobile sensors to characterize the human behavior and develop a new generation of CAPTCHAs. The experiments are evaluated with HuMIdb Human Mobile Interaction database, a novel multimodal mobile database that comprises 14 mobile sensors acquired from 600 users. HuMIdb is freely available to the research community. |