Datasets:
sswt
/
arxiv_sample_50K

Dataset card Data Studio Files
xet
Community
1
text
stringlengths
62
2.94k
Compositional Exploration of Combinatorial Scientific Models ; We implement a novel representation of model search spaces as diagrams over a category of models, where we have restricted attention to a broad class of models whose structure is presented by Csets. Colimits in these diagram categories allow the creation of composite model spaces from more primitive spaces. We present a novel implementation of the computer algebra of finitely presented categories and diagram categories including their limits and colimits, which formalizes a notion of model space exploration. This is coupled with strategies to facilitate the selection of desired models from these model spaces. We demonstrate our framework by generating a tool which fits experimental data, searching an epidemiologyrelevant subspace of massaction kinetic models.
Green Runner A tool for efficient model selection from model repositories ; Deep learning models have become essential in software engineering, enabling intelligent features like image captioning and document generation. However, their popularity raises concerns about environmental impact and inefficient model selection. This paper introduces GreenRunnerGPT, a novel tool for efficiently selecting deep learning models based on specific use cases. It employs a large language model to suggest weights for quality indicators, optimizing resource utilization. The tool utilizes a multiarmed bandit framework to evaluate models against target datasets, considering tradeoffs. We demonstrate that GreenRunnerGPT is able to identify a model suited to a target use case without wasteful computations that would occur under a bruteforce approach to model selection.
General Design Bayesian Generalized Linear Mixed Models ; Linear mixed models are able to handle an extraordinary range of complications in regressiontype analyses. Their most common use is to account for withinsubject correlation in longitudinal data analysis. They are also the standard vehicle for smoothing spatial count data. However, when treated in full generality, mixed models can also handle splinetype smoothing and closely approximate kriging. This allows for nonparametric regression models e.g., additive models and varying coefficient models to be handled within the mixed model framework. The key is to allow the random effects design matrix to have general structure; hence our label general design. For continuous response data, particularly when Gaussianity of the response is reasonably assumed, computation is now quite mature and supported by the R, SAS and SPLUS packages. Such is not the case for binary and count responses, where generalized linear mixed models GLMMs are required, but are hindered by the presence of intractable multivariate integrals. Software known to us supports special cases of the GLMM e.g., PROC NLMIXED in SAS or glmmML in R or relies on the sometimes crude Laplacetype approximation of integrals e.g., the SAS macro glimmix or glmmPQL in R. This paper describes the fitting of general design generalized linear mixed models. A Bayesian approach is taken and Markov chain Monte Carlo MCMC is used for estimation and inference. In this generalized setting, MCMC requires sampling from nonstandard distributions. In this article, we demonstrate that the MCMC package WinBUGS facilitates sound fitting of general design Bayesian generalized linear mixed models in practice.
On the Asymptotic Behaviour of Cosmological Models in ScalarTensor Theories of Gravity ; We study the qualitative properties of cosmological models in scalartensor theories of gravity by exploiting the formal equivalence of these theories with general relativity minimally coupled to a scalar field under a conformal transformation and field redefinition. In particular, we investigate the asymptotic behaviour of spatially homogeneous cosmological models in a class of scalartensor theories which are conformally equivalent to general relativistic Bianchi cosmologies with a scalar field and an exponential potential whose qualitative features have been studied previously. Particular attention is focussed on those scalartensor theory cosmological models, which are shown to be selfsimilar, that correspond to general relativistic models that play an important role in describing the asymptotic behaviour of more general models e.g., those cosmological models that act as earlytime and latetime attractors.
CP Violation and Family Mixing in the Effective Electroweak Lagrangian ; We construct the most general effective Lagrangian of the matter sector of the Standard Model, including mixing and CP violating terms. The Lagrangian contains the effective operators that give the leading contribution in theories where the physics beyond the Standard Model shows at a scale Lambda MW. We perform the diagonalization and passage to the physical basis in full generality. We determine the contribution to the different observables and discuss the possible new sources of CP violation, the idea being to be able to gain some knowledge about new physics beyond the Standard Model from general considerations, without having to compute model by model. The values of the coefficients of the effective Lagrangian in some theories, including the Standard Model, are presented and we try to draw some general conclusions about the general pattern exhibited by physics beyond the Standard Model in what concerns CP violation. In the process we have had to deal with two theoretical problems which are very interesting in their own the renormalization of the CKM matrix elements and the wave function renormalization in the onshell scheme when mixing is present.
Exploring Social Influence for Recommendation A Probabilistic Generative Model Approach ; In this paper, we propose a probabilistic generative model, called unified model, which naturally unifies the ideas of social influence, collaborative filtering and contentbased methods for item recommendation. To address the issue of hidden social influence, we devise new algorithms to learn the model parameters of our proposal based on expectation maximization EM. In addition to a singlemachine version of our EM algorithm, we further devise a parallelized implementation on the MapReduce framework to process two largescale datasets we collect. Moreover, we show that the social influence obtained from our generative models can be used for group recommendation. Finally, we conduct comprehensive experiments using the datasets crawled from last.fm and whrrl.com to validate our ideas. Experimental results show that the generative models with social influence significantly outperform those without incorporating social influence. The unified generative model proposed in this paper obtains the best performance. Moreover, our study on social influence finds that users in whrrl.com are more likely to get influenced by friends than those in last.fm. The experimental results also confirm that our social influence based group recommendation algorithm outperforms the stateoftheart algorithms for group recommendation.
Generating Functionals of Random Packing Point Processes From HardCore to Carrier Sensing ; In this paper we study the generating functionals of several random packing processes the classical Mat'ern hardcore model; its extensions, the kMat'ern models and the inftyMat'ern model, which is an example of random sequential packing process. We first give a sufficient condition for the inftyMat'ern model to be welldefined unlike the other two, the latter may not be welldefined on unbounded spaces. Then the generating functional of the resulting point process is given for each of the three models as the solution of a differential equation. Series representations and bounds on the generating functional of the packing models are also derived. Last but not least, we obtain moment measures and Palm distributions of the considered packing models departing from their generating functionals.
General Intensity Shapes in Optimal Liquidation ; The classical literature on optimal liquidation, rooted in AlmgrenChriss models, tackles the optimal liquidation problem using a tradeoff between market impact and price risk. Therefore, it only answers the general question of the optimal liquidation rhythm. The very question of the actual way to proceed with liquidation is then rarely dealt with. Our model, that incorporates both price risk and nonexecution risk, is an attempt to tackle this question using limit orders. The very general framework we propose to model liquidation generalizes the existing literature on optimal posting of limit orders. We consider a riskadverse agent whereas the model of Bayraktar and Ludkovski only tackles the case of a riskneutral one. We consider very general functional forms for the execution process intensity, whereas Gu'eant et al. is restricted to exponential intensity. Eventually, we link the execution cost function of AlmgrenChriss models to the intensity function in our model, providing then a way to see AlmgrenChriss models as a limit of ours.
Towards a LogicBased Unifying Framework for Computing ; In this paper we propose a logicbased, framework inspired by artificial intelligence, but scaled down for practical database and programming applications. Computation in the framework is viewed as the task of generating a sequence of state transitions, with the purpose of making an agent's goals all true. States are represented by sets of atomic sentences or facts, representing the values of program variables, tuples in a coordination language, facts in relational databases, or Herbrand models. In the modeltheoretic semantics, the entire sequence of states and events are combined into a single modeltheoretic structure, by associating timestamps with facts and events. But in the operational semantics, facts are updated destructively, without timestamps. We show that the model generated by destructive updates is identical to the model generated by reasoning with facts containing timestamps. We also extend the model with intentional predicates and composite event predicates defined by logic programs containing conditions in firstorder logic, which query the current state.
Exploring the foundations of quantum mechanics using Monte Carlo simulations of the FreedmanClauser experimental test of Bell's Inequality ; Monte Carlo simulations of the FreedmanClauser experiment are used to test the generic wave function collapse model of Quantum Mechanics, a local realistic model, and a dynamical state reduction model of wave function collapse. The simulated results are compared to the actual results of the experiment which confirmed the quantum mechanical calculation for nine different relative angles between the two polarization analyzers. For each simulation 5times107 total simulated photon pairs were generated at each relative angle. The generic wave function collapse model closely followed the general shape of the theoretical calculation but differed from the calculated values by 2.5 to 3.3 for angles less than or equal to pi8 and differed by 15.0 to 52.5 for angles greater than or equal to 3pi8. The local realistic model did not replicate the experimental results but was generally within 1 of a classical calculation for all analyzer angles. A dynamical state reductioncollapse model, approximated by using a smeared polarization, yielded values within 1 of the quantum mechanical calculation and provides an independent estimate of the correlation length used in these models of rc 1.04 pm 0.14times 105 cm.
A note on the evaluation of generative models ; Probabilistic generative models can be used for compression, denoising, inpainting, texture synthesis, semisupervised learning, unsupervised feature learning, and other tasks. Given this wide range of applications, it is not surprising that a lot of heterogeneity exists in the way these models are formulated, trained, and evaluated. As a consequence, direct comparison between models is often difficult. This article reviews mostly known but often underappreciated properties relating to the evaluation and interpretation of generative models with a focus on image models. In particular, we show that three of the currently most commonly used criteriaaverage loglikelihood, Parzen window estimates, and visual fidelity of samplesare largely independent of each other when the data is highdimensional. Good performance with respect to one criterion therefore need not imply good performance with respect to the other criteria. Our results show that extrapolation from one criterion to another is not warranted and generative models need to be evaluated directly with respect to the applications they were intended for. In addition, we provide examples demonstrating that Parzen window estimates should generally be avoided.
Deep Reinforcement Learning for Dialogue Generation ; Recent neural models of dialogue generation offer great promise for generating responses for conversational agents, but tend to be shortsighted, predicting utterances one at a time while ignoring their influence on future outcomes. Modeling the future direction of a dialogue is crucial to generating coherent, interesting dialogues, a need which led traditional NLP models of dialogue to draw on reinforcement learning. In this paper, we show how to integrate these goals, applying deep reinforcement learning to model future reward in chatbot dialogue. The model simulates dialogues between two virtual agents, using policy gradient methods to reward sequences that display three useful conversational properties informativity nonrepetitive turns, coherence, and ease of answering related to forwardlooking function. We evaluate our model on diversity, length as well as with human judges, showing that the proposed algorithm generates more interactive responses and manages to foster a more sustained conversation in dialogue simulation. This work marks a first step towards learning a neural conversational model based on the longterm success of dialogues.
Implicit Tubular Surface Generation Guided by Centerline ; Most machine learningbased coronary artery segmentation methods represent the vascular lumen surface in an implicit way by the centerline and the associated lumen radii, which makes the subsequent modeling process to generate a whole piece of watertight coronary artery tree model difficult. To solve this problem, in this paper, we propose a modeling method with the learningbased segmentation results by 1 considering mesh vertices as physical particles and using interaction force model and particle expansion model to generate uniformly distributed point cloud on the implicit lumen surface and; 2 doing incremental Delaunaybased triangulation. Our method has the advantage of being able to consider the complex shape of the coronary artery tree as a whole piece; hence no extra stitching or intersection removal algorithm is needed to generate a watertight model. Experiment results demonstrate that our method is capable of generating high quality mesh model which is highly consistent with the given implicit vascular lumen surface, with an average error of 0.08 mm.
Generalized CurieWeiss Model and Quadratic Pressure in Ergodic Theory ; We explain the Curie Weiss model in Statistical Mechanics within the Ergodic viewpoint. More precisely, we simultaneously define in 1,1mathbbN, on the one hand a generalized Curie Weiss model within the Statistical Mechanics viewpoint and on the other hand, quadratic free energy and quadratic pressure within the Ergodic Theory viewpoint. We show that there are finitely many invariant measures which maximize the quadratic free energy. They are all Dynamical Gibbs Measures. Moreover, the Probabilistic Gibbs measures for generalized Curie Weiss model converge to a determined combination of the dynamical conformal measures associated to these Dynamical Gibbs Measures. The standard Curie Weiss model is a particular case of our generalized Curie Weiss model. An Ergodic viewpoint over the Curie Weiss Potts model is also given.
Text to Image Synthesis Using Generative Adversarial Networks ; Generating images from natural language is one of the primary applications of recent conditional generative models. Besides testing our ability to model conditional, highly dimensional distributions, text to image synthesis has many exciting and practical applications such as photo editing or computeraided content creation. Recent progress has been made using Generative Adversarial Networks GANs. This material starts with a gentle introduction to these topics and discusses the existent state of the art models. Moreover, I propose Wasserstein GANCLS, a new model for conditional image generation based on the Wasserstein distance which offers guarantees of stability. Then, I show how the novel loss function of Wasserstein GANCLS can be used in a Conditional Progressive Growing GAN. In combination with the proposed loss, the model boosts by 7.07 the best Inception Score on the Caltech birds dataset of the models which use only the sentencelevel visual semantics. The only model which performs better than the Conditional Wasserstein Progressive Growing GAN is the recently proposed AttnGAN which uses wordlevel visual semantics as well.
On Visual Hallmarks of Robustness to Adversarial Malware ; A central challenge of adversarial learning is to interpret the resulting hardened model. In this contribution, we ask how robust generalization can be visually discerned and whether a concise view of the interactions between a hardened decision map and input samples is possible. We first provide a means of visually comparing a hardened model's loss behavior with respect to the adversarial variants generated during training versus loss behavior with respect to adversarial variants generated from other sources. This allows us to confirm that the association of observed flatness of a loss landscape with generalization that is seen with naturally trained models extends to adversarially hardened models and robust generalization. To complement these means of interpreting model parameter robustness we also use selforganizing maps to provide a visual means of superimposing adversarial and natural variants on a model's decision space, thus allowing the model's global robustness to be comprehensively examined.
Structured Generative Models of Natural Source Code ; We study the problem of building generative models of natural source code NSC; that is, source code written and understood by humans. Our primary contribution is to describe a family of generative models for NSC that have three key properties First, they incorporate both sequential and hierarchical structure. Second, we learn a distributed representation of source code elements. Finally, they integrate closely with a compiler, which allows leveraging compiler logic and abstractions when building structure into the model. We also develop an extension that includes more complex structure, refining how the model generates identifier tokens based on what variables are currently in scope. Our models can be learned efficiently, and we show empirically that including appropriate structure greatly improves the models, measured by the probability of generating test programs.
A generic model for spouse's pensions with a view towards the calculation of liabilities ; We introduce a generic model for spouse's pensions. The generic model allows for the modeling of various types of spouse's pensions with payments commencing at the death of the insured. We derive abstract formulas for cashflows and liabilities corresponding to common types of spouse's pensions. We show how the standard formulas from the Danish G82 concession can be obtained as a special case of our generic model. We also derive expressions for liabilities for spouse's pensions in models more advanced than found in the G82 concession. The generic nature of our model and results furthermore enable the calculation of cashflows and liabilities using simple estimates of marital behaviour among a population.
Learning Deep Generative Spatial Models for Mobile Robots ; We propose a new probabilistic framework that allows mobile robots to autonomously learn deep, generative models of their environments that span multiple levels of abstraction. Unlike traditional approaches that combine engineered models for lowlevel features, geometry, and semantics, our approach leverages recent advances in SumProduct Networks SPNs and deep learning to learn a single, universal model of the robot's spatial environment. Our model is fully probabilistic and generative, and represents a joint distribution over spatial information ranging from lowlevel geometry to semantic interpretations. Once learned, it is capable of solving a wide range of tasks from semantic classification of places, uncertainty estimation, and novelty detection, to generation of place appearances based on semantic information and prediction of missing data in partial observations. Experiments on laserrange data from a mobile robot show that the proposed universal model obtains performance superior to stateoftheart models finetuned to one specific task, such as Generative Adversarial Networks GANs or SVMs.
Speakerindependent raw waveform model for glottal excitation ; Recent speech technology research has seen a growing interest in using WaveNets as statistical vocoders, i.e., generating speech waveforms from acoustic features. These models have been shown to improve the generated speech quality over classical vocoders in many tasks, such as texttospeech synthesis and voice conversion. Furthermore, conditioning WaveNets with acoustic features allows sharing the waveform generator model across multiple speakers without additional speaker codes. However, multispeaker WaveNet models require large amounts of training data and computation to cover the entire acoustic space. This paper proposes leveraging the sourcefilter model of speech production to more effectively train a speakerindependent waveform generator with limited resources. We present a multispeaker 'GlotNet' vocoder, which utilizes a WaveNet to generate glottal excitation waveforms, which are then used to excite the corresponding vocal tract filter to produce speech. Listening tests show that the proposed model performs favourably to a direct WaveNet vocoder trained with the same model architecture and data.
Adversarial Training of Variational Autoencoders for High Fidelity Image Generation ; Variational autoencoders VAEs provide an attractive solution to image generation problem. However, they tend to produce blurred and oversmoothed images due to their dependence on pixelwise reconstruction loss. This paper introduces a new approach to alleviate this problem in the VAE based generative models. Our model simultaneously learns to match the data, reconstruction loss and the latent distributions of real and fake images to improve the quality of generated samples. To compute the loss distributions, we introduce an autoencoder based discriminator model which allows an adversarial learning procedure. The discriminator in our model also provides perceptual guidance to the VAE by matching the learned similarity metric of the real and fake samples in the latent space. To stabilize the overall training process, our model uses an error feedback approach to maintain the equilibrium between competing networks in the model. Our experiments show that the generated samples from our proposed model exhibit a diverse set of attributes and facial expressions and scale up to highresolution images very well.
ATMAdversarialneural Topic Model ; Topic models are widely used for thematic structure discovery in text. But traditional topic models often require dedicated inference procedures for specific tasks at hand. Also, they are not designed to generate wordlevel semantic representations. To address these limitations, we propose a topic modeling approach based on Generative Adversarial Nets GANs, called Adversarialneural Topic Model ATM. The proposed ATM models topics with Dirichlet prior and employs a generator network to capture the semantic patterns among latent topics. Meanwhile, the generator could also produce wordlevel semantic representations. To illustrate the feasibility of porting ATM to tasks other than topic modeling, we apply ATM for open domain event extraction. Our experimental results on the two public corpora show that ATM generates more coherence topics, outperforming a number of competitive baselines. Moreover, ATM is able to extract meaningful events from news articles.
Auditing Data Provenance in TextGeneration Models ; To help enforce dataprotection regulations such as GDPR and detect unauthorized uses of personal data, we develop a new emphmodel auditing technique that helps users check if their data was used to train a machine learning model. We focus on auditing deeplearning models that generate naturallanguage text, including word prediction and dialog generation. These models are at the core of popular online services and are often trained on personal data such as users' messages, searches, chats, and comments. We design and evaluate a blackbox auditing method that can detect, with very few queries to a model, if a particular user's texts were used to train it among thousands of other users. We empirically show that our method can successfully audit wellgeneralized models that are not overfitted to the training data. We also analyze how textgeneration models memorize word sequences and explain why this memorization makes them amenable to auditing.
Learning a Generator Model from Terminal Bus Data ; In this work we investigate approaches to reconstruct generator models from measurements available at the generator terminal bus using machine learning ML techniques. The goal is to develop an emulator which is trained online and is capable of fast predictive computations. The training is illustrated on synthetic data generated based on available opensource dynamical generator model. Two ML techniques were developed and tested a standard vector autoregressive VAR model; and b novel customized long shortterm memory LSTM deep learning model. Tradeoffs in reconstruction ability between computationally light but linear AR model and powerful but computationally demanding LSTM model are established and analyzed.
Density Forecasts and the Leverage Effect Some Evidence from Observation and ParameterDriven Volatility Models ; The leverage effect refers to the wellestablished relationship between returns and volatility. When returns fall, volatility increases. We examine the role of the leverage effect with regards to generating density forecasts of equity returns using wellknown observation and parameterdriven volatility models. These models differ in their assumptions regarding The parametric specification, the evolution of the conditional volatility process and how the leverage effect is accounted for. The ability of a model to generate accurate density forecasts when the leverage effect is incorporated or not as well as a comparison between different modeltypes is carried out using a large number of financial timeseries. We find that, models with the leverage effect generally generate more accurate density forecasts compared to their noleverage counterparts. Moreover, we also find that our choice with regards to how to model the leverage effect and the conditional logvolatility process is important in generating accurate density forecasts
Deep Quantization Encoding Convolutional Activations with Deep Generative Model ; Deep convolutional neural networks CNNs have proven highly effective for visual recognition, where learning a universal representation from activations of convolutional layer plays a fundamental problem. In this paper, we present Fisher Vector encoding with Variational AutoEncoder FVVAE, a novel deep architecture that quantizes the local activations of convolutional layer in a deep generative model, by training them in an endtoend manner. To incorporate FV encoding strategy into deep generative models, we introduce Variational AutoEncoder model, which steers a variational inference and learning in a neural network which can be straightforwardly optimized using standard stochastic gradient method. Different from the FV characterized by conventional generative models e.g., Gaussian Mixture Model which parsimoniously fit a discrete mixture model to data distribution, the proposed FVVAE is more flexible to represent the natural property of data for better generalization. Extensive experiments are conducted on three public datasets, i.e., UCF101, ActivityNet, and CUB2002011 in the context of video action recognition and finegrained image classification, respectively. Superior results are reported when compared to stateoftheart representations. Most remarkably, our proposed FVVAE achieves todate the best published accuracy of 94.2 on UCF101.
Adjusted quantile residual for generalized linear models ; Generalized linear models are widely used in many areas of knowledge. As in other classes of regression models, it is desirable to perform diagnostic analysis in generalized linear models using residuals that are approximately standard normally distributed. Diagnostic analysis in this class of models are usually performed using the standardized Pearson residual or the standardized deviance residual. The former has skewed distribution and the latter has negative mean, specially when the variance of the response variable is high. In this work, we introduce the adjusted quantile residual for generalized linear models. Using Monte Carlo simulation techniques and two applications, we compare this residual with the standardized Pearson residual, the standardized deviance residual and two other residuals. Overall, the results suggest that the adjusted quantile residual is a better tool for diagnostic analysis in generalized linear models.
Exploring cooperative game mechanisms of scientific coauthorship networks ; Scientific coauthorship, generated by collaborations and competitions among researchers, reflects effective organizations of human resources. Researchers, their expected benefits through collaborations, and their cooperative costs constitute the elements of a game. Hence we propose a cooperative game model to explore the evolution mechanisms of scientific coauthorship networks. The model generates geometric hypergraphs, where the costs are modelled by space distances, and the benefits are expressed by node reputations, i. e. geometric zones that depend on node position in space and time. Modelled cooperative strategies conditioned on positive benefitminuscost reflect the spatial reciprocity principle in collaborations, and generate high clustering and degree assortativity, two typical features of coauthorship networks. Modelled reputations generate the generalized Poisson parts and fat tails appeared in specific distributions of empirical data, e. g. paper team size distribution. The combined effect of modelled costs and reputations reproduces the transitions emerged in degree distribution, in the correlation between degree and local clustering coefficient, etc. The model provides an example of how individual strategies induce network complexity, as well as an application of game theory to social affiliation networks.
Query and Output Generating Words by Querying Distributed Word Representations for Paraphrase Generation ; Most recent approaches use the sequencetosequence model for paraphrase generation. The existing sequencetosequence model tends to memorize the words and the patterns in the training dataset instead of learning the meaning of the words. Therefore, the generated sentences are often grammatically correct but semantically improper. In this work, we introduce a novel model based on the encoderdecoder framework, called Word Embedding Attention Network WEAN. Our proposed model generates the words by querying distributed word representations i.e. neural word embeddings, hoping to capturing the meaning of the according words. Following previous work, we evaluate our model on two paraphraseoriented tasks, namely text simplification and short text abstractive summarization. Experimental results show that our model outperforms the sequencetosequence baseline by the BLEU score of 6.3 and 5.5 on two English text simplification datasets, and the ROUGE2 F1 score of 5.7 on a Chinese summarization dataset. Moreover, our model achieves stateoftheart performances on these three benchmark datasets.
Understanding the Limitations of Conditional Generative Models ; Classconditional generative models hold promise to overcome the shortcomings of their discriminative counterparts. They are a natural choice to solve discriminative tasks in a robust manner as they jointly optimize for predictive performance and accurate modeling of the input distribution. In this work, we investigate robust classification with likelihoodbased generative models from a theoretical and practical perspective to investigate if they can deliver on their promises. Our analysis focuses on a spectrum of robustness properties 1 Detection of worstcase outliers in the form of adversarial examples; 2 Detection of averagecase outliers in the form of ambiguous inputs and 3 Detection of incorrectly labeled indistribution inputs. Our theoretical result reveals that it is impossible to guarantee detectability of adversariallyperturbed inputs even for nearoptimal generative classifiers. Experimentally, we find that while we are able to train robust models for MNIST, robustness completely breaks down on CIFAR10. We relate this failure to various undesirable model properties that can be traced to the maximum likelihood training objective. Despite being a common choice in the literature, our results indicate that likelihoodbased conditional generative models may are surprisingly ineffective for robust classification.
Bias Correction of Learned Generative Models using LikelihoodFree Importance Weighting ; A learned generative model often produces biased statistics relative to the underlying data distribution. A standard technique to correct this bias is importance sampling, where samples from the model are weighted by the likelihood ratio under model and true distributions. When the likelihood ratio is unknown, it can be estimated by training a probabilistic classifier to distinguish samples from the two distributions. We employ this likelihoodfree importance weighting method to correct for the bias in generative models. We find that this technique consistently improves standard goodnessoffit metrics for evaluating the sample quality of stateoftheart deep generative models, suggesting reduced bias. Finally, we demonstrate its utility on representative applications in a data augmentation for classification using generative adversarial networks, and b modelbased policy evaluation using offpolicy data.
FMRI data augmentation via synthesis ; We present an empirical evaluation of fMRI data augmentation via synthesis. For synthesis we use generative models trained on real neuroimaging data to produce novel taskdependent functional brain images. Analyzed generative models include classic approaches such as the Gaussian mixture model GMM, and modern implicit generative models such as the generative adversarial network GAN and the variational autoencoder VAE. In particular, the proposed GAN and VAE models utilize 3dimensional convolutions, which enables modeling of highdimensional brain image tensors with structured spatial correlations. The synthesized datasets are then used to augment classifiers designed to predict cognitive and behavioural outcomes. Our results suggest that the proposed models are able to generate highquality synthetic brain images which are diverse and taskdependent. Perhaps most importantly, the performance improvements of data augmentation via synthesis are shown to be complementary to the choice of the predictive model. Thus, our results suggest that data augmentation via synthesis is a promising approach to address the limited availability of fMRI data, and to improve the quality of predictive fMRI models.
Semantic Pyramid for Image Generation ; We present a novel GANbased model that utilizes the space of deep features learned by a pretrained classification model. Inspired by classical image pyramid representations, we construct our model as a Semantic Generation Pyramid a hierarchical framework which leverages the continuum of semantic information encapsulated in such deep features; this ranges from low level information contained in fine features to high level, semantic information contained in deeper features. More specifically, given a set of features extracted from a reference image, our model generates diverse image samples, each with matching features at each semantic level of the classification model. We demonstrate that our model results in a versatile and flexible framework that can be used in various classic and novel image generation tasks. These include generating images with a controllable extent of semantic similarity to a reference image, and different manipulation tasks such as semanticallycontrolled inpainting and compositing; all achieved with the same model, with no further training.
Envelopes of equivalent martingale measures and a generalized noarbitrage principle in a finite setting ; We consider a oneperiod market model composed by a riskfree asset and a risky asset with n possible future values namely, a nnomial market model. We characterize the lower envelope of the class of equivalent martingale measures in such market model, showing that it is a belief function, obtained as the strict convex combination of two necessity measures. Then, we reformulate a general oneperiod pricing problem in the framework of belief functions this allows to model frictions in the market and can be justified in terms of partially resolving uncertainty according to Jaffray. We provide a generalized noarbitrage condition for a generic oneperiod market model under partially resolving uncertainty and show that the riskneutral belief function arising in the oneperiod nnomial market model does not satisfy such condition. Finally, we derive a generalized arbitragefree lower pricing rule through an inner approximation of the riskneutral belief function arising in the oneperiod nnomial market model.
Deep Generative Models with Learnable Knowledge Constraints ; The broad set of deep generative models DGMs has achieved remarkable advances. However, it is often difficult to incorporate rich structured domain knowledge with the endtoend DGMs. Posterior regularization PR offers a principled framework to impose structured constraints on probabilistic models, but has limited applicability to the diverse DGMs that can lack a Bayesian formulation or even explicit density evaluation. PR also requires constraints to be fully specified a priori, which is impractical or suboptimal for complex knowledge with learnable uncertain parts. In this paper, we establish mathematical correspondence between PR and reinforcement learning RL, and, based on the connection, expand PR to learn constraints as the extrinsic reward in RL. The resulting algorithm is modelagnostic to apply to any DGMs, and is flexible to adapt arbitrary constraints with the model jointly. Experiments on human image generation and templated sentence generation show models with learned knowledge constraints by our algorithm greatly improve over base generative models.
A Generalized Kinetic Model for Heterogeneous GasSolid Reactions ; We present a generalized kinetic model for gassolid heterogeneous reactions taking place at the interface between two phases. The model studies the reaction kinetics by taking into account the reactions at the interface, as well as the transport process within the product layer. The standard unreacted shrinking core model relies on the assumption of quasistatic diffusion that results in a steadystate concentration profile of gas reactant in the product layer. By relaxing this assumption and resolving the entire problem, general solutions can be obtained for reaction kinetics, including the reaction front velocity and the conversion volume fraction of reacted solid. The unreacted shrinking core model is shown to be accurate and in agreement with the generalized model for slow reaction or fast diffusion, low concentration of gas reactant, and small solid size. Otherwise, a generalized kinetic model should be used.
The Importance of Generation Order in Language Modeling ; Neural language models are a critical component of stateoftheart systems for machine translation, summarization, audio transcription, and other tasks. These language models are almost universally autoregressive in nature, generating sentences one token at a time from left to right. This paper studies the influence of token generation order on model quality via a novel twopass language model that produces partiallyfilled sentence templates and then fills in missing tokens. We compare various strategies for structuring these two passes and observe a surprisingly large variation in model quality. We find the most effective strategy generates function words in the first pass followed by content words in the second. We believe these experimental results justify a more extensive investigation of generation order for neural language models.
Unsupervised Domain Adaptation using Generative Models and Selfensembling ; Transferring knowledge across different datasets is an important approach to successfully train deep models with a smallscale target dataset or when few labeled instances are available. In this paper, we aim at developing a model that can generalize across multiple domain shifts, so that this model can adapt from a single source to multiple targets. This can be achieved by randomizing the generation of the data of various styles to mitigate the domain mismatch. First, we present a new adaptation to the CycleGAN model to produce stochastic style transfer between two image batches of different domains. Second, we enhance the classifier performance by using a selfensembling technique with a teacher and student model to train on both original and generated data. Finally, we present experimental results on three datasets Office31, OfficeHome, and Visual Domain adaptation. The results suggest that selfensembling is better than simple data augmentation with the newly generated data and a single model trained this way can have the best performance across all different transfer tasks.
Approximate Query Processing using Deep Generative Models ; Data is generated at an unprecedented rate surpassing our ability to analyze them. The database community has pioneered many novel techniques for Approximate Query Processing AQP that could give approximate results in a fraction of time needed for computing exact results. In this work, we explore the usage of deep learning DL for answering aggregate queries specifically for interactive applications such as data exploration and visualization. We use deep generative models, an unsupervised learning based approach, to learn the data distribution faithfully such that aggregate queries could be answered approximately by generating samples from the learned model. The model is often compact few hundred KBs so that arbitrary AQP queries could be answered on the client side without contacting the database server. Our other contributions include identifying model bias and minimizing it through a rejection sampling based approach and an algorithm to build model ensembles for AQP for improved accuracy. Our extensive experiments show that our proposed approach can provide answers with high accuracy and low latency.
Do Massively Pretrained Language Models Make Better Storytellers ; Large neural language models trained on massive amounts of text have emerged as a formidable strategy for Natural Language Understanding tasks. However, the strength of these models as Natural Language Generators is less clear. Though anecdotal evidence suggests that these models generate better quality text, there has been no detailed study characterizing their generation abilities. In this work, we compare the performance of an extensively pretrained model, OpenAI GPT2117 Radford et al., 2019, to a stateoftheart neural story generation model Fan et al., 2018. By evaluating the generated text across a wide variety of automatic metrics, we characterize the ways in which pretrained models do, and do not, make better storytellers. We find that although GPT2117 conditions more strongly on context, is more sensitive to ordering of events, and uses more unusual words, it is just as likely to produce repetitive and underdiverse text when using likelihoodmaximizing decoding algorithms.
ImpedanceBased WholeSystem Modeling for a Composite Grid via FrameDynamics Embedding ; The paper establishes a methodology to overcome the difficulty of dynamic frame alignment and system separation in impedance modeling of ac grids, and thereby enables impedancebased wholesystem modeling of generatorconverter composite power systems. The methodology is based on a framedynamicsembedding transformation via an intermediary steady frame between local and global frames, which yields a locally defined impedance model for each generator or converter that does not rely on a global frame but retains all frame dynamics. The individual impedance model can then be readily combined into a wholesystem model even for meshed networks via the proposed closedloop formulation without network separation. Compared to startoftheart impedancebased models, the proposed method retains both frame dynamics and scalability, and is generally applicable to various network topologies meshed, radial, etc and combinations of machines generators, motors, converters, etc. The methodology is used to analyze the dynamic interaction between generators and converters in a composite grid, which yields important findings and potential solutions for unstable oscillation caused by PLLswing coupling in lowinertia grids.
Improving Grammatical Error Correction with Machine Translation Pairs ; We propose a novel data synthesis method to generate diverse errorcorrected sentence pairs for improving grammatical error correction, which is based on a pair of machine translation models of different qualities i.e., poor and good. The poor translation model resembles the ESL English as a second language learner and tends to generate translations of low quality in terms of fluency and grammatical correctness, while the good translation model generally generates fluent and grammatically correct translations. We build the poor and good translation model with phrasebased statistical machine translation model with decreased language model weight and neural machine translation model respectively. By taking the pair of their translations of the same sentences in a bridge language as errorcorrected sentence pairs, we can construct unlimited pseudo parallel data. Our approach is capable of generating diverse fluencyimproving patterns without being limited by the predefined rule set and the seed errorcorrected data. Experimental results demonstrate the effectiveness of our approach and show that it can be combined with other synthetic data sources to yield further improvements.
Implicit Generative Modeling for Efficient Exploration ; Efficient exploration remains a challenging problem in reinforcement learning, especially for those tasks where rewards from environments are sparse. A commonly used approach for exploring such environments is to introduce some intrinsic reward. In this work, we focus on model uncertainty estimation as an intrinsic reward for efficient exploration. In particular, we introduce an implicit generative modeling approach to estimate a Bayesian uncertainty of the agent's belief of the environment dynamics. Each random draw from our generative model is a neural network that instantiates the dynamic function, hence multiple draws would approximate the posterior, and the variance in the future prediction based on this posterior is used as an intrinsic reward for exploration. We design a training algorithm for our generative model based on the amortized Stein Variational Gradient Descent. In experiments, we compare our implementation with stateoftheart intrinsic rewardbased exploration approaches, including two recent approaches based on an ensemble of dynamic models. In challenging exploration tasks, our implicit generative model consistently outperforms competing approaches regarding data efficiency in exploration.
Distributional Invariances and Interventional Markov Equivalence for Mixed Graph Models ; The invariance properties of interventional distributions relative to the observational distribution, and how these properties allow us to refine Markov equivalence classes MECs of DAGs, is central to causal DAG discovery algorithms that use both interventional and observational data. Here, we show how the invariance properties of interventional DAG models, and the corresponding refinement of MECs into interventional MECs, can be generalized to mixed graphical models that allow for latent cofounders and selection variables. We first generalize interventional Markov equivalence to all formal independence models associated to loopless mixed graphs. For ancestral graphs, we prove the resulting interventional MECs admit a graphical characterization generalizing that of DAGs. We then define interventional distributions for acyclic directed mixed graph models, and prove that this generalization aligns with the graphical generalization of interventional Markov equivalence given for the formal independence models. This provides a framework for causal model discovery via observational and interventional data in the presence of latent confounders that applies even when the interventions are uncontrolled.
A Paraconsistent ASPlike Language with Tractable Model Generation ; Answer Set Programming ASP is nowadays a dominant rulebased knowledge representation tool. Though existing ASP variants enjoy efficient implementations, generating an answer set remains intractable. The goal of this research is to define a new asplike rule language, 4SP, with tractable model generation. The language combines ideas of ASP and a paraconsistent rule language 4QL. Though 4SP shares the syntax of asp and for each program all its answer sets are among 4SP models, the new language differs from ASP in its logical foundations, the intended methodology of its use and complexity of computing models. As we show in the paper, 4QL can be seen as a paraconsistent counterpart of ASP programs stratified with respect to default negation. Although model generation of wellsupported models for 4QL programs is tractable, dropping stratification makes both 4QL and ASP intractable. To retain tractability while allowing nonstratified programs, in 4SP we introduce trial expressions interlacing programs with hypotheses as to the truth values of default negations. This allows us to develop amodel generation algorithm with deterministic polynomial time complexity. We also show relationships among 4SP, ASP and 4QL.
Rethinking Generalization of Neural Models A Named Entity Recognition Case Study ; While neural networkbased models have achieved impressive performance on a large body of NLP tasks, the generalization behavior of different models remains poorly understood Does this excellent performance imply a perfect generalization model, or are there still some limitations In this paper, we take the NER task as a testbed to analyze the generalization behavior of existing models from different perspectives and characterize the differences of their generalization abilities through the lens of our proposed measures, which guides us to better design models and training methods. Experiments with indepth analyses diagnose the bottleneck of existing neural NER models in terms of breakdown performance analysis, annotation errors, dataset bias, and category relationships, which suggest directions for improvement. We have released the datasets ReCoNLL, PLONER for the future research at our project page httppfliu.comInterpretNER. As a byproduct of this paper, we have opensourced a project that involves a comprehensive summary of recent NER papers and classifies them into different research topics httpsgithub.compfliunlpNamedEntityRecognitionNERPapers.
GlowTTS A Generative Flow for TexttoSpeech via Monotonic Alignment Search ; Recently, texttospeech TTS models such as FastSpeech and ParaNet have been proposed to generate melspectrograms from text in parallel. Despite the advantage, the parallel TTS models cannot be trained without guidance from autoregressive TTS models as their external aligners. In this work, we propose GlowTTS, a flowbased generative model for parallel TTS that does not require any external aligner. By combining the properties of flows and dynamic programming, the proposed model searches for the most probable monotonic alignment between text and the latent representation of speech on its own. We demonstrate that enforcing hard monotonic alignments enables robust TTS, which generalizes to long utterances, and employing generative flows enables fast, diverse, and controllable speech synthesis. GlowTTS obtains an orderofmagnitude speedup over the autoregressive model, Tacotron 2, at synthesis with comparable speech quality. We further show that our model can be easily extended to a multispeaker setting.
GoalDirected Planning for Habituated Agents by Active Inference Using a Variational Recurrent Neural Network ; It is crucial to ask how agents can achieve goals by generating action plans using only partial models of the world acquired through habituated sensorymotor experiences. Although many existing robotics studies use a forward model framework, there are generalization issues with high degrees of freedom. The current study shows that the predictive coding PC and active inference AIF frameworks, which employ a generative model, can develop better generalization by learning a prior distribution in a low dimensional latent state space representing probabilistic structures extracted from well habituated sensorymotor trajectories. In our proposed model, learning is carried out by inferring optimal latent variables as well as synaptic weights for maximizing the evidence lower bound, while goaldirected planning is accomplished by inferring latent variables for maximizing the estimated lower bound. Our proposed model was evaluated with both simple and complex robotic tasks in simulation, which demonstrated sufficient generalization in learning with limited training data by setting an intermediate value for a regularization coefficient. Furthermore, comparative simulation results show that the proposed model outperforms a conventional forward model in goaldirected planning, due to the learned prior confining the search of motor plans within the range of habituated trajectories.
Tractably Modelling Dependence in Networks Beyond Exchangeability ; We propose a general framework for modelling network data that is designed to describe aspects of nonexchangeable networks. Conditional on latent unobserved variables, the edges of the network are generated by their finite growth history with latent orders while the marginal probabilities of the adjacency matrix are modeled by a generalization of a graph limit function or a graphon. In particular, we study the estimation, clustering and degree behavior of the network in our setting. We determine i the minimax estimator of a composite graphon with respect to squared error loss; ii that spectral clustering is able to consistently detect the latent membership when the blockwise constant composite graphon is considered under additional conditions; and iii we are able to construct models with heavytailed empirical degrees under specific scenarios and parameter choices. This explores why and under which general conditions nonexchangeable network data can be described by a stochastic block model. The new modelling framework is able to capture empirically important characteristics of network data such as sparsity combined with heavy tailed degree distribution, and add understanding as to what generative mechanisms will make them arise. Keywords statistical network analysis, exchangeable arrays, stochastic block model, nonlinear stochastic processes.
Unsupervised Generative Adversarial Alignment Representation for Sheet music, Audio and Lyrics ; Sheet music, audio, and lyrics are three main modalities during writing a song. In this paper, we propose an unsupervised generative adversarial alignment representation UGAAR model to learn deep discriminative representations shared across three major musical modalities sheet music, lyrics, and audio, where a deep neural network based architecture on three branches is jointly trained. In particular, the proposed model can transfer the strong relationship between audio and sheet music to audiolyrics and sheetlyrics pairs by learning the correlation in the latent shared subspace. We apply CCA components of audio and sheet music to establish new ground truth. The generative G model learns the correlation of two couples of transferred pairs to generate new audiosheet pair for a fixed lyrics to challenge the discriminative D model. The discriminative model aims at distinguishing the input which is from the generative model or the ground truth. The two models simultaneously train in an adversarial way to enhance the ability of deep alignment representation learning. Our experimental results demonstrate the feasibility of our proposed UGAAR for alignment representation learning among sheet music, audio, and lyrics.
Learning to Generate Music With Sentiment ; Deep Learning models have shown very promising results in automatically composing polyphonic music pieces. However, it is very hard to control such models in order to guide the compositions towards a desired goal. We are interested in controlling a model to automatically generate music with a given sentiment. This paper presents a generative Deep Learning model that can be directed to compose music with a given sentiment. Besides music generation, the same model can be used for sentiment analysis of symbolic music. We evaluate the accuracy of the model in classifying sentiment of symbolic music using a new dataset of video game soundtracks. Results show that our model is able to obtain good prediction accuracy. A user study shows that human subjects agreed that the generated music has the intended sentiment, however negative pieces can be ambiguous.
GLM General Language Model Pretraining with Autoregressive Blank Infilling ; There have been various types of pretraining architectures including autoencoding models e.g., BERT, autoregressive models e.g., GPT, and encoderdecoder models e.g., T5. However, none of the pretraining frameworks performs the best for all tasks of three main categories including natural language understanding NLU, unconditional generation, and conditional generation. We propose a General Language Model GLM based on autoregressive blank infilling to address this challenge. GLM improves blank filling pretraining by adding 2D positional encodings and allowing an arbitrary order to predict spans, which results in performance gains over BERT and T5 on NLU tasks. Meanwhile, GLM can be pretrained for different types of tasks by varying the number and lengths of blanks. On a wide range of tasks across NLU, conditional and unconditional generation, GLM outperforms BERT, T5, and GPT given the same model sizes and data, and achieves the best performance from a single pretrained model with 1.25x parameters of BERT Large , demonstrating its generalizability to different downstream tasks.
Graphbased Normalizing Flow for Human Motion Generation and Reconstruction ; Datadriven approaches for modeling human skeletal motion have found various applications in interactive media and social robotics. Challenges remain in these fields for generating highfidelity samples and robustly reconstructing motion from imperfect input data, due to e.g. missed marker detection. In this paper, we propose a probabilistic generative model to synthesize and reconstruct long horizon motion sequences conditioned on past information and control signals, such as the path along which an individual is moving. Our method adapts the existing work MoGlow by introducing a new graphbased model. The model leverages the spatialtemporal graph convolutional network STGCN to effectively capture the spatial structure and temporal correlation of skeletal motion data at multiple scales. We evaluate the models on a mixture of motion capture datasets of human locomotion with footstep and bonelength analysis. The results demonstrate the advantages of our model in reconstructing missing markers and achieving comparable results on generating realistic future poses. When the inputs are imperfect, our model shows improvements on robustness of generation.
ScoreGrad Multivariate Probabilistic Time Series Forecasting with Continuous Energybased Generative Models ; Multivariate time series prediction has attracted a lot of attention because of its wide applications such as intelligence transportation, AIOps. Generative models have achieved impressive results in time series modeling because they can model data distribution and take noise into consideration. However, many existing works can not be widely used because of the constraints of functional form of generative models or the sensitivity to hyperparameters. In this paper, we propose ScoreGrad, a multivariate probabilistic time series forecasting framework based on continuous energybased generative models. ScoreGrad is composed of time series feature extraction module and conditional stochastic differential equation based score matching module. The prediction can be achieved by iteratively solving reversetime SDE. To the best of our knowledge, ScoreGrad is the first continuous energy based generative model used for time series forecasting. Furthermore, ScoreGrad achieves stateoftheart results on six realworld datasets. The impact of hyperparameters and sampler types on the performance are also explored. Code is available at httpsgithub.comyantijinScoreGradPred.
Abstraction of Markov Population Dynamics via Generative Adversarial Nets ; Markov Population Models are a widespread formalism used to model the dynamics of complex systems, with applications in Systems Biology and many other fields. The associated Markov stochastic process in continuous time is often analyzed by simulation, which can be costly for large or stiff systems, particularly when a massive number of simulations has to be performed e.g. in a multiscale model. A strategy to reduce computational load is to abstract the population model, replacing it with a simpler stochastic model, faster to simulate. Here we pursue this idea, building on previous works and constructing a generator capable of producing stochastic trajectories in continuous space and discrete time. This generator is learned automatically from simulations of the original model in a Generative Adversarial setting. Compared to previous works, which rely on deep neural networks and Dirichlet processes, we explore the use of state of the art generative models, which are flexible enough to learn a full trajectory rather than a single transition kernel.
Generating the Graph Gestalt KernelRegularized Graph Representation Learning ; Recent work on graph generative models has made remarkable progress towards generating increasingly realistic graphs, as measured by global graph features such as degree distribution, density, and clustering coefficients. Deep generative models have also made significant advances through better modelling of the local correlations in the graph topology, which have been very useful for predicting unobserved graph components, such as the existence of a link or the class of a node, from nearby observed graph components. A complete scientific understanding of graph data should address both global and local structure. In this paper, we propose a joint model for both as complementary objectives in a graph VAE framework. Global structure is captured by incorporating graph kernels in a probabilistic model whose loss function is closely related to the maximum mean discrepancyMMD between the global structures of the reconstructed and the input graphs. The ELBO objective derived from the model regularizes a standard local link reconstruction term with an MMD term. Our experiments demonstrate a significant improvement in the realism of the generated graph structures, typically by 12 orders of magnitude of graph structure metrics, compared to leading graph VAEand GAN models. Local link reconstruction improves as well in many cases.
Image2Lego Customized LEGO Set Generation from Images ; Although LEGO sets have entertained generations of children and adults, the challenge of designing customized builds matching the complexity of realworld or imagined scenes remains too great for the average enthusiast. In order to make this feat possible, we implement a system that generates a LEGO brick model from 2D images. We design a novel solution to this problem that uses an octreestructured autoencoder trained on 3D voxelized models to obtain a feasible latent representation for model reconstruction, and a separate network trained to predict this latent representation from 2D images. LEGO models are obtained by algorithmic conversion of the 3D voxelized model to bricks. We demonstrate firstofitskind conversion of photographs to 3D LEGO models. An octree architecture enables the flexibility to produce multiple resolutions to best fit a user's creative vision or design needs. In order to demonstrate the broad applicability of our system, we generate stepbystep building instructions and animations for LEGO models of objects and human faces. Finally, we test these automatically generated LEGO sets by constructing physical builds using real LEGO bricks.
SideControl Controlled Opendomain Dialogue Generation via Additive Side Networks ; Transformerbased pretrained language models boost the performance of opendomain dialogue systems. Prior works leverage Transformerbased pretrained language models to generate texts with desired attributes in two general approaches 1 gradientbased methods updating all latent representations of pretrained models with gradients from attribute models; 2 weighteddecoding methods reranking beam candidates from pretrained models with attribute functions. However, gradientbased methods lead to high computation cost and can easily get overfitted on small training sets, while weighteddecoding methods are inherently constrained by the lowvariance highbias pretrained model. In this work, we propose a novel approach to control the generation of Transformerbased pretrained language models the SideControl framework, which leverages a novel control attributes loss to incorporate useful control signals, and is shown to perform well with very limited training samples. We evaluate our proposed method on two benchmark opendomain dialogue datasets, and results show that the SideControl framework has better controllability, higher generation quality and better sampleefficiency than existing gradientbased and weighteddecoding baselines.
Nonparametric Functional Analysis of Generalized Linear Models Under Nonlinear Constraints ; This article introduces a novel nonparametric methodology for Generalized Linear Models which combines the strengths of the binary regression and latent variable formulations for categorical data, while overcoming their disadvantages. Requiring minimal assumptions, it extends recently published parametric versions of the methodology and generalizes it. If the underlying data generating process is asymmetric, it gives uniformly better prediction and inference performance over the parametric formulation. Furthermore, it introduces a new classification statistic utilizing which I show that overall, it has better model fit, inference and classification performance than the parametric version, and the difference in performance is statistically significant especially if the data generating process is asymmetric. In addition, the methodology can be used to perform model diagnostics for any model specification. This is a highly useful result, and it extends existing work for categorical model diagnostics broadly across the sciences. The mathematical results also highlight important new findings regarding the interplay of statistical significance and scientific significance. Finally, the methodology is applied to various realworld datasets to show that it may outperform widely used existing models, including Random Forests and Deep Neural Networks with very few iterations.
Response Generation with ContextAware Prompt Learning ; Pretrained language models PLM have marked a huge leap in neural dialogue modeling. While PLMs are pretrained on largescale text corpora, they are usually finetuned on scarce dialogue data with specific domain knowledge and dialogue styles. However, tailoring the language models while fully utilizing prior knowledge in large pretrained models remains a challenge. In this paper, we present a novel approach for pretrained dialogue modeling that casts the dialogue generation problem as a promptlearning task. Instead of finetuning on limited dialogue data, our approach, DialogPrompt, learns continuous prompt embeddings optimized for dialogue contexts, which appropriately elicit knowledge from the large pretrained model. To encourage the model to better utilize the prompt embeddings, the prompt encoders are designed to be dynamically generated based on the dialogue context. Experiments on popular conversation datasets show that our approach significantly outperforms the finetuning baseline and the generic promptlearning methods. Furthermore, human evaluations strongly support the superiority of DialogPrompt in regard to response generation quality.
FairStyle Debiasing StyleGAN2 with Style Channel Manipulations ; Recent advances in generative adversarial networks have shown that it is possible to generate highresolution and hyperrealistic images. However, the images produced by GANs are only as fair and representative as the datasets on which they are trained. In this paper, we propose a method for directly modifying a pretrained StyleGAN2 model that can be used to generate a balanced set of images with respect to one e.g., eyeglasses or more attributes e.g., gender and eyeglasses. Our method takes advantage of the style space of the StyleGAN2 model to perform disentangled control of the target attributes to be debiased. Our method does not require training additional models and directly debiases the GAN model, paving the way for its use in various downstream applications. Our experiments show that our method successfully debiases the GAN model within a few minutes without compromising the quality of the generated images. To promote fair generative models, we share the code and debiased models at httpcatlabteam.github.iofairstyle.
Participation FactorBased Adaptive Model Reduction for Fast Power System Simulation ; This paper describes an adaptive method to reduce a nonlinear power system model for fast and accurate transient stability simulation. It presents an approach to analyze and rank participation factors of each system state variable into dominant system modes excited by a disturbance so as to determine which regions or generators can be reduced without impacting the accuracy of simulation for a study area. In this approach, the generator models located in an external area with large participation factors are nonlinearly reduced and the rest of the generators will be linearized. The simulation results confirm that the assessment of the level of interaction between generators and system modes by participation factors is effective in enhancing the accuracy and speed of power system models. The proposed method is applied to the Northeastern Power Coordinating Council region system with 48machine, 140bus power system model and the results are compared with two cases including fully linearized model reduction and model reduction using the rotor angle deviation criteria.
A ModelAgnostic Data Manipulation Method for Personabased Dialogue Generation ; Towards building intelligent dialogue agents, there has been a growing interest in introducing explicit personas in generation models. However, with limited personabased dialogue data at hand, it may be difficult to train a dialogue generation model well. We point out that the data challenges of this generation task lie in two aspects first, it is expensive to scale up current personabased dialogue datasets; second, each data sample in this task is more complex to learn with than conventional dialogue data. To alleviate the above data issues, we propose a data manipulation method, which is modelagnostic to be packed with any personabased dialogue generation model to improve its performance. The original training samples will first be distilled and thus expected to be fitted more easily. Next, we show various effective ways that can diversify such easier distilled data. A given base model will then be trained via the constructed data curricula, i.e. first on augmented distilled samples and then on original ones. Experiments illustrate the superiority of our method with two strong base dialogue models Transformer encoderdecoder and GPT2.
DialogVED A Pretrained Latent Variable EncoderDecoder Model for Dialog Response Generation ; Dialog response generation in open domain is an important research topic where the main challenge is to generate relevant and diverse responses. In this paper, we propose a new dialog pretraining framework called DialogVED, which introduces continuous latent variables into the enhanced encoderdecoder pretraining framework to increase the relevance and diversity of responses. With the help of a large dialog corpus Reddit, we pretrain the model using the following 4 tasks adopted in language models LMs and variational autoencoders VAEs 1 masked language model; 2 response generation; 3 bagofwords prediction; and 4 KL divergence reduction. We also add additional parameters to model the turn structure in dialogs to improve the performance of the pretrained model. We conduct experiments on PersonaChat, DailyDialog, and DSTC7AVSD benchmarks for response generation. Experimental results show that our model achieves the new stateoftheart results on all these datasets.
Dynamic Domain Generalization ; Domain generalization DG is a fundamental yet very challenging research topic in machine learning. The existing arts mainly focus on learning domaininvariant features with limited source domains in a static model. Unfortunately, there is a lack of trainingfree mechanism to adjust the model when generalized to the agnostic target domains. To tackle this problem, we develop a brandnew DG variant, namely Dynamic Domain Generalization DDG, in which the model learns to twist the network parameters to adapt the data from different domains. Specifically, we leverage a metaadjuster to twist the network parameters based on the static model with respect to different data from different domains. In this way, the static model is optimized to learn domainshared features, while the metaadjuster is designed to learn domainspecific features. To enable this process, DomainMix is exploited to simulate data from diverse domains during teaching the metaadjuster to adapt to the upcoming agnostic target domains. This learning mechanism urges the model to generalize to different agnostic target domains via adjusting the model without training. Extensive experiments demonstrate the effectiveness of our proposed method. Code is available at httpsgithub.comMetaVisionLabDDG
Human Biophysics as Network Weights Conditional Generative Models for Dynamic Simulation ; Simulations of biophysical systems are fundamental for studying physiological mechanisms and developing human machine interfaces. Whilst advanced numerical methods, such as finite element models, can excel in this task, they are extremely computationally expensive to use when generating a large number of simulations or simulating dynamic events with continuously changing structural parameters. We propose an architecture that uses a conditional generative model to interpolate between the numerical model states, dramatically lowering the modeling time while maintaining a high generation accuracy. As a demonstration of this concept, we present BioMime, a hybridstructured generative model that enables an accurate, ultrafast, and arbitrarily high temporalresolution simulation of a specific biophysical system during dynamic changes. This methodology has wide applications in physiological and clinical research as well as in supporting data augmentation strategies for signal analysis, representing a computationally efficient and highly accurate model for biophysical simulations.
I Hear Your True Colors Image Guided Audio Generation ; We propose Im2Wav, an image guided opendomain audio generation system. Given an input image or a sequence of images, Im2Wav generates a semantically relevant sound. Im2Wav is based on two Transformer language models, that operate over a hierarchical discrete audio representation obtained from a VQVAE based model. We first produce a lowlevel audio representation using a language model. Then, we upsample the audio tokens using an additional language model to generate a highfidelity audio sample. We use the rich semantics of a pretrained CLIP Contrastive LanguageImage Pretraining embedding as a visual representation to condition the language model. In addition, to steer the generation process towards the conditioning image, we apply the classifierfree guidance method. Results suggest that Im2Wav significantly outperforms the evaluated baselines in both fidelity and relevance evaluation metrics. Additionally, we provide an ablation study to better assess the impact of each of the method components on overall performance. Lastly, to better evaluate imagetoaudio models, we propose an outofdomain image dataset, denoted as ImageHear. ImageHear can be used as a benchmark for evaluating future imagetoaudio models. Samples and code can be found inside the manuscript.
On the Compositional Generalization Gap of InContext Learning ; Pretrained large generative language models have shown great performance on many tasks, but exhibit low compositional generalization abilities. Scaling such models has been shown to improve their performance on various NLP tasks even just by conditioning them on a few examples to solve the task without any finetuning also known as incontext learning. In this work, we look at the gap between the indistribution ID and outofdistribution OOD performance of such models in semantic parsing tasks with incontext learning. In the ID settings, the demonstrations are from the same split test or train that the model is being evaluated on, and in the OOD settings, they are from the other split. We look at how the relative generalization gap of incontext learning evolves as models are scaled up. We evaluate four model families, OPT, BLOOM, CodeGen and Codex on three semantic parsing datasets, CFQ, SCAN and GeoQuery with different number of exemplars, and observe a trend of decreasing relative generalization gap as models are scaled up.
GAMMT Generative Ambiguity Modeling Using Multiple Transformers ; We introduce a novel model called GAMMT Generative Ambiguity Models using Multiple Transformers for sequential data that is based on sets of probabilities. Unlike conventional models, our approach acknowledges that the data generation process of a sequence is not deterministic, but rather ambiguous and influenced by a set of probabilities. To capture this ambiguity, GAMMT employs multiple parallel transformers that are linked by a selection mechanism, allowing for the approximation of ambiguous probabilities. The generative nature of our approach also enables multiple representations of input tokens and sequences. While our models have not yet undergone experimental validation, we believe that our model has great potential to achieve high quality and diversity in modeling sequences with uncertain data generation processes.
Coder Reviewer Reranking for Code Generation ; Sampling diverse programs from a code language model and reranking with model likelihood is a popular method for code generation but it is prone to preferring degenerate solutions. Inspired by collaborative programming, we propose CoderReviewer reranking. We augment Coder language models from past work, which generate programs given language instructions, with Reviewer models, which evaluate the likelihood of the instruction given the generated programs. We perform an extensive study across six datasets with eight models from three model families. Experimental results show that CoderReviewer reranking leads to consistent and significant improvement up to 17 absolute accuracy gain over reranking with the Coder model only. When combined with executability filtering, CoderReviewer reranking can often outperform the minimum Bayes risk method. CoderReviewer reranking is easy to implement by prompting, can generalize to different programming languages, and works well with offtheshelf hyperparameters.
Diffusion Generative Models in Infinite Dimensions ; Diffusion generative models have recently been applied to domains where the available data can be seen as a discretization of an underlying function, such as audio signals or time series. However, these models operate directly on the discretized data, and there are no semantics in the modeling process that relate the observed data to the underlying functional forms. We generalize diffusion models to operate directly in function space by developing the foundational theory for such models in terms of Gaussian measures on Hilbert spaces. A significant benefit of our function space point of view is that it allows us to explicitly specify the space of functions we are working in, leading us to develop methods for diffusion generative modeling in Sobolev spaces. Our approach allows us to perform both unconditional and conditional generation of functionvalued data. We demonstrate our methods on several synthetic and realworld benchmarks.
Local Limit of Nonlocal Gravity A Teleparallel Extension of General Relativity ; We describe a general constitutive framework for a teleparallel extension of the general theory of relativity. This approach goes beyond the teleparallel equivalent of general relativity TEGR by broadening the analogy with the electrodynamics of media. In particular, the main purpose of this paper is to investigate in detail a local constitutive extension of TEGR that is the local limit of nonlocal gravity NLG. Within this framework, we study the modified FLRW cosmological models. Of these, the most cogent turns out to be the modified flat model which is shown to be inconsistent with the existence of a positive cosmological constant. Moreover, dynamic dark energy and other components of the modified flat model evolve differently with the expansion of the universe as compared to the standard flat cosmological model. The observational consequences of the modified flat model are briefly explored and it is shown that the model is capable of resolving the H0 tension.
Geometryaware Autoregressive Models for Calorimeter Shower Simulations ; Calorimeter shower simulations are often the bottleneck in simulation time for particle physics detectors. A lot of effort is currently spent on optimizing generative architectures for specific detector geometries, which generalize poorly. We develop a geometryaware autoregressive model on a range of calorimeter geometries such that the model learns to adapt its energy deposition depending on the size and position of the cells. This is a key proofofconcept step towards building a model that can generalize to new unseen calorimeter geometries with little to no additional training. Such a model can replace the hundreds of generative models used for calorimeter simulation in a Large Hadron Collider experiment. For the study of future detectors, such a model will dramatically reduce the large upfront investment usually needed to generate simulations.
Generalized Standard Model with higherorder derivatives under Rotor Mechanism and its Quantization ; The Standard Model is the paradigm of particle physics which gives an accurate theory for fundamental particle interactions. However, the extension of Standard Model with higherorder derivatives is not a wellstudied subject. This paper is a followup work of the previous study of the generalized Abelian gauge field theory and YangMills theory under rotor mechanism of order n of higher order derivatives, and we apply it to the Standard Model of particle physics. Rotor mechanism on scalar field and Dirac field is also studied. We will study the quantization of the rotored Standard Model using path integral approach. We also inherit the previous result from the path integral quantization of generalized Abelian gauge field and apply it to our nonAbelian case. Then we carry out the generalized BRST quantization and prove the existence of the SlavnovTaylor Identities of the rotor model. Finally, we discuss the possibility of rotor model on taming the infinities arise from the selfenergy correction of the Higgs boson in high spacetime dimension, thus this provides a partial solution and new insights to the Hierarchy problem.
Versatile EnergyBased Probabilistic Models for High Energy Physics ; As a classical generative modeling approach, energybased models have the natural advantage of flexibility in the form of the energy function. Recently, energybased models have achieved great success in modeling highdimensional data in computer vision and natural language processing. In line with these advancements, we build a multipurpose energybased probabilistic model for High Energy Physics events at the Large Hadron Collider. This framework builds on a powerful generative model and describes higherorder interparticle interactions.It suits different encoding architectures and builds on implicit generation. As for applicational aspects, it can serve as a powerful parameterized event generator for physics simulation, a generic anomalous signal detector free from spurious correlations, and an augmented event classifier for particle identification.
Generative Machine Learning for Detector Response Modeling with a Conditional Normalizing Flow ; In this paper, we explore the potential of generative machine learning models as an alternative to the computationally expensive Monte Carlo MC simulations commonly used by the Large Hadron Collider LHC experiments. Our objective is to develop a generative model capable of efficiently simulating detector responses for specific particle observables, focusing on the correlations between detector responses of different particles in the same event and accommodating asymmetric detector responses. We present a conditional normalizing flow model CNF based on a chain of Masked Autoregressive Flows, which effectively incorporates conditional variables and models highdimensional density distributions. We assess the performance of the cnf model using a simulated sample of Higgs boson decaying to diphoton events at the LHC. We create reconstructionlevel observables using a smearing technique. We show that conditional normalizing flows can accurately model complex detector responses and their correlation. This method can potentially reduce the computational burden associated with generating large numbers of simulated events while ensuring that the generated events meet the requirements for data analyses.
DDRF Denoising Diffusion Model for Remote Sensing Image Fusion ; Denosing diffusion model, as a generative model, has received a lot of attention in the field of image generation recently, thanks to its powerful generation capability. However, diffusion models have not yet received sufficient research in the field of image fusion. In this article, we introduce diffusion model to the image fusion field, treating the image fusion task as imagetoimage translation and designing two different conditional injection modulation modules i.e., style transfer modulation and wavelet modulation to inject coarsegrained style information and finegrained highfrequency and lowfrequency information into the diffusion UNet, thereby generating fused images. In addition, we also discussed the residual learning and the selection of training objectives of the diffusion model in the image fusion task. Extensive experimental results based on quantitative and qualitative assessments compared with benchmarks demonstrates stateoftheart results and good generalization performance in image fusion tasks. Finally, it is hoped that our method can inspire other works and gain insight into this field to better apply the diffusion model to image fusion tasks. Code shall be released for better reproducibility.
Posttraining Model Quantization Using GANs for Synthetic Data Generation ; Quantization is a widely adopted technique for deep neural networks to reduce the memory and computational resources required. However, when quantized, most models would need a suitable calibration process to keep their performance intact, which requires data from the target domain, such as a fraction of the dataset used in model training and model validation i.e. calibration dataset. In this study, we investigate the use of synthetic data as a substitute for the calibration with real data for the quantization method. We propose a data generation method based on Generative Adversarial Networks that are trained prior to the model quantization step. We compare the performance of models quantized using data generated by StyleGAN2ADA and our pretrained DiStyleGAN, with quantization using real data and an alternative data generation method based on fractal images. Overall, the results of our experiments demonstrate the potential of leveraging synthetic data for calibration during the quantization process. In our experiments, the percentage of accuracy degradation of the selected models was less than 0.6, with our best performance achieved on MobileNetV2 0.05. The code is available at httpsgithub.comThanosM97gsoc2022openvino
Analyzing Bias in Diffusionbased Face Generation Models ; Diffusion models are becoming increasingly popular in synthetic data generation and image editing applications. However, these models can amplify existing biases and propagate them to downstream applications. Therefore, it is crucial to understand the sources of bias in their outputs. In this paper, we investigate the presence of bias in diffusionbased face generation models with respect to attributes such as gender, race, and age. Moreover, we examine how dataset size affects the attribute composition and perceptual quality of both diffusion and Generative Adversarial Network GAN based face generation models across various attribute classes. Our findings suggest that diffusion models tend to worsen distribution bias in the training data for various attributes, which is heavily influenced by the size of the dataset. Conversely, GAN models trained on balanced datasets with a larger number of samples show less bias across different attributes.
TwoinOne A Model Hijacking Attack Against Text Generation Models ; Machine learning has progressed significantly in various applications ranging from face recognition to text generation. However, its success has been accompanied by different attacks. Recently a new attack has been proposed which raises both accountability and parasitic computing risks, namely the model hijacking attack. Nevertheless, this attack has only focused on image classification tasks. In this work, we broaden the scope of this attack to include text generation and classification models, hence showing its broader applicability. More concretely, we propose a new model hijacking attack, Ditto, that can hijack different text classification tasks into multiple generation ones, e.g., language translation, text summarization, and language modeling. We use a range of text benchmark datasets such as SST2, TweetEval, AGnews, QNLI, and IMDB to evaluate the performance of our attacks. Our results show that by using Ditto, an adversary can successfully hijack text generation models without jeopardizing their utility.
Data Redaction from Conditional Generative Models ; Deep generative models are known to produce undesirable samples such as harmful content. Traditional mitigation methods include retraining from scratch, filtering, or editing; however, these are either computationally expensive or can be circumvented by third parties. In this paper, we take a different approach and study how to postedit an alreadytrained conditional generative model so that it redacts certain conditionals that will, with high probability, lead to undesirable content. This is done by distilling the conditioning network in the models, giving a solution that is effective, efficient, controllable, and universal for a class of deep generative models. We conduct experiments on redacting prompts in texttoimage models and redacting voices in texttospeech models. Our method is computationally light, leads to better redaction quality and robustness than baseline methods while still retaining high generation quality.
AudioToken Adaptation of TextConditioned Diffusion Models for AudiotoImage Generation ; In recent years, image generation has shown a great leap in performance, where diffusion models play a central role. Although generating highquality images, such models are mainly conditioned on textual descriptions. This begs the question how can we adopt such models to be conditioned on other modalities. In this paper, we propose a novel method utilizing latent diffusion models trained for texttoimagegeneration to generate images conditioned on audio recordings. Using a pretrained audio encoding model, the proposed method encodes audio into a new token, which can be considered as an adaptation layer between the audio and text representations. Such a modeling paradigm requires a small number of trainable parameters, making the proposed approach appealing for lightweight optimization. Results suggest the proposed method is superior to the evaluated baseline methods, considering objective and subjective metrics. Code and samples are available at httpspages.cs.huji.ac.iladiyosslabAudioToken.
Are Diffusion Models VisionAndLanguage Reasoners ; Textconditioned image generation models have recently shown immense qualitative success using denoising diffusion processes. However, unlike discriminative visionandlanguage models, it is a nontrivial task to subject these diffusionbased generative models to automatic finegrained quantitative evaluation of highlevel phenomena such as compositionality. Towards this goal, we perform two innovations. First, we transform diffusionbased models in our case, Stable Diffusion for any imagetext matching ITM task using a novel method called DiffusionITM. Second, we introduce the GenerativeDiscriminative Evaluation Benchmark GDBench benchmark with 7 complex visionandlanguage tasks, bias evaluation and detailed analysis. We find that Stable Diffusion DiffusionITM is competitive on many tasks and outperforms CLIP on compositional tasks like like CLEVR and Winoground. We further boost its compositional performance with a transfer setup by finetuning on MSCOCO while retaining generative capabilities. We also measure the stereotypical bias in diffusion models, and find that Stable Diffusion 2.1 is, for the most part, less biased than Stable Diffusion 1.5. Overall, our results point in an exciting direction bringing discriminative and generative model evaluation closer. We will release code and benchmark setup soon.
Fewshot Finetuning vs. Incontext Learning A Fair Comparison and Evaluation ; Fewshot finetuning and incontext learning are two alternative strategies for task adaptation of pretrained language models. Recently, incontext learning has gained popularity over finetuning due to its simplicity and improved outofdomain generalization, and because extensive evidence shows that finetuned models pick up on spurious correlations. Unfortunately, previous comparisons of the two approaches were done using models of different sizes. This raises the question of whether the observed weaker outofdomain generalization of finetuned models is an inherent property of finetuning or a limitation of the experimental setup. In this paper, we compare the generalization of fewshot finetuning and incontext learning to challenge datasets, while controlling for the models used, the number of examples, and the number of parameters, ranging from 125M to 30B. Our results show that finetuned language models can in fact generalize well outofdomain. We find that both approaches generalize similarly; they exhibit large variation and depend on properties such as model size and the number of examples, highlighting that robust task adaptation remains a challenge.
Unsupervised Anomaly Detection in Medical Images Using Masked Diffusion Model ; It can be challenging to identify brain MRI anomalies using supervised deeplearning techniques due to anatomical heterogeneity and the requirement for pixellevel labeling. Unsupervised anomaly detection approaches provide an alternative solution by relying only on samplelevel labels of healthy brains to generate a desired representation to identify abnormalities at the pixel level. Although, generative models are crucial for generating such anatomically consistent representations of healthy brains, accurately generating the intricate anatomy of the human brain remains a challenge. In this study, we present a method called maskedDDPM mDPPM, which introduces maskingbased regularization to reframe the generation task of diffusion models. Specifically, we introduce Masked Image Modeling MIM and Masked Frequency Modeling MFM in our selfsupervised approach that enables models to learn visual representations from unlabeled data. To the best of our knowledge, this is the first attempt to apply MFM in DPPM models for medical applications. We evaluate our approach on datasets containing tumors and numerous sclerosis lesions and exhibit the superior performance of our unsupervised method as compared to the existing fullyweakly supervised baselines. Code is available at httpsgithub.comhasan1292mDDPM.
Few Shot Rationale Generation using SelfTraining with Dual Teachers ; Selfrationalizing models that also generate a freetext explanation for their predicted labels are an important tool to build trustworthy AI applications. Since generating explanations for annotated labels is a laborious and costly pro cess, recent models rely on large pretrained language models PLMs as their backbone and fewshot learning. In this work we explore a selftraining approach leveraging both labeled and unlabeled data to further improve fewshot models, under the assumption that neither human written rationales nor annotated task labels are available at scale. We introduce a novel dualteacher learning framework, which learns two specialized teacher models for task prediction and rationalization using selftraining and distills their knowledge into a multitasking student model that can jointly generate the task label and rationale. Furthermore, we formulate a new loss function, Masked Label Regularization MLR which promotes explanations to be strongly conditioned on predicted labels. Evaluation on three public datasets demonstrate that the proposed methods are effective in modeling task labels and generating faithful rationales.
Stochastic MultiPerson 3D Motion Forecasting ; This paper aims to deal with the ignored realworld complexities in prior work on human motion forecasting, emphasizing the social properties of multiperson motion, the diversity of motion and social interactions, and the complexity of articulated motion. To this end, we introduce a novel task of stochastic multiperson 3D motion forecasting. We propose a duallevel generative modeling framework that separately models independent individual motion at the local level and social interactions at the global level. Notably, this duallevel modeling mechanism can be achieved within a shared generative model, through introducing learnable latent codes that represent intents of future motion and switching the codes' modes of operation at different levels. Our framework is general; we instantiate it with different generative models, including generative adversarial networks and diffusion models, and various multiperson forecasting models. Extensive experiments on CMUMocap, MuPoTS3D, and SoMoF benchmarks show that our approach produces diverse and accurate multiperson predictions, significantly outperforming the state of the art.
Face0 Instantaneously Conditioning a TexttoImage Model on a Face ; We present Face0, a novel way to instantaneously condition a texttoimage generation model on a face, in sample time, without any optimization procedures such as finetuning or inversions. We augment a dataset of annotated images with embeddings of the included faces and train an image generation model, on the augmented dataset. Once trained, our system is practically identical at inference time to the underlying base model, and is therefore able to generate images, given a usersupplied face image and a prompt, in just a couple of seconds. Our method achieves pleasing results, is remarkably simple, extremely fast, and equips the underlying model with new capabilities, like controlling the generated images both via text or via direct manipulation of the input face embeddings. In addition, when using a fixed random vector instead of a face embedding from a user supplied image, our method essentially solves the problem of consistent character generation across images. Finally, while requiring further research, we hope that our method, which decouples the model's textual biases from its biases on faces, might be a step towards some mitigation of biases in future texttoimage models.
Latent Dynamical Implicit Diffusion Processes ; Latent dynamical models are commonly used to learn the distribution of a latent dynamical process that represents a sequence of noisy data samples. However, producing samples from such models with high fidelity is challenging due to the complexity and variability of latent and observation dynamics. Recent advances in diffusionbased generative models, such as DDPM and NCSN, have shown promising alternatives to stateoftheart latent generative models, such as Neural ODEs, RNNs, and Normalizing flow networks, for generating highquality sequential samples from a prior distribution. However, their application in modeling sequential data with latent dynamical models is yet to be explored. Here, we propose a novel latent variable model named latent dynamical implicit diffusion processes LDIDPs, which utilizes implicit diffusion processes to sample from dynamical latent processes and generate sequential observation samples accordingly. We tested LDIDPs on synthetic and simulated neural decoding problems. We demonstrate that LDIDPs can accurately learn the dynamics over latent dimensions. Furthermore, the implicit sampling method allows for the computationally efficient generation of highquality sequential data samples from the latent and observation spaces.
Neural Embeddings for Web Testing ; Web test automation techniques employ web crawlers to automatically produce a web app model that is used for test generation. Existing crawlers rely on appspecific, thresholdbased, algorithms to assess state equivalence. Such algorithms are hard to tune in the general case and cannot accurately identify and remove nearduplicate web pages from crawl models. Failing to retrieve an accurate web app model results in automated test generation solutions that produce redundant test cases and inadequate test suites that do not cover the web app functionalities adequately. In this paper, we propose WEBEMBED, a novel abstraction function based on neural network embeddings and thresholdfree classifiers that can be used to produce accurate web app models during modelbased test generation. Our evaluation on nine web apps shows that WEBEMBED outperforms stateoftheart techniques by detecting nearduplicates more accurately, inferring better web app models that exhibit 22 more precision, and 24 more recall on average. Consequently, the test suites generated from these models achieve higher code coverage, with improvements ranging from 2 to 59 on an appwise basis and averaging at 23.
BigWavGAN A WaveToWave Generative Adversarial Network for Music SuperResolution ; Generally, Deep Neural Networks DNNs are expected to have high performance when their model size is large. However, large models failed to produce highquality results commensurate with their scale in music SuperResolution SR. We attribute this to that DNNs cannot learn information commensurate with their size from standard mean square error losses. To unleash the potential of large DNN models in music SR, we propose BigWavGAN, which incorporates Demucs, a largescale wavetowave model, with StateOfTheArt SOTA discriminators and adversarial training strategies. Our discriminator consists of MultiScale Discriminator MSD and MultiResolution Discriminator MRD. During inference, since only the generator is utilized, there are no additional parameters or computational resources required compared to the baseline model Demucs. Objective evaluation affirms the effectiveness of BigWavGAN in music SR. Subjective evaluations indicate that BigWavGAN can generate music with significantly high perceptual quality over the baseline model. Notably, BigWavGAN surpasses the SOTA music SR model in both simulated and realworld scenarios. Moreover, BigWavGAN represents its superior generalization ability to address outofdistribution data. The conducted ablation study reveals the importance of our discriminators and training strategies. Samples are available on the demo page.
Fair GANs through model rebalancing with synthetic data ; Deep generative models require large amounts of training data. This often poses a problem as the collection of datasets can be expensive and difficult, in particular datasets that are representative of the appropriate underlying distribution e.g. demographic. This introduces biases in datasets which are further propagated in the models. We present an approach to mitigate biases in an existing generative adversarial network by rebalancing the model distribution. We do so by generating balanced data from an existing unbalanced deep generative model using latent space exploration and using this data to train a balanced generative model. Further, we propose a bias mitigation loss function that shows improvements in the fairness metric even when trained with unbalanced datasets. We show results for the Stylegan2 models while training on the FFHQ dataset for racial fairness and see that the proposed approach improves on the fairness metric by almost 5 times, whilst maintaining image quality. We further validate our approach by applying it to an imbalanced Cifar10 dataset. Lastly, we argue that the traditionally used image quality metrics such as Frechet inception distance FID are unsuitable for bias mitigation problems.
Let Quantum Neural Networks Choose Their Own Frequencies ; Parameterized quantum circuits as machine learning models are typically well described by their representation as a partial Fourier series of the input features, with frequencies uniquely determined by the feature map's generator Hamiltonians. Ordinarily, these dataencoding generators are chosen in advance, fixing the space of functions that can be represented. In this work we consider a generalization of quantum models to include a set of trainable parameters in the generator, leading to a trainable frequency TF quantum model. We numerically demonstrate how TF models can learn generators with desirable properties for solving the task at hand, including nonregularly spaced frequencies in their spectra and flexible spectral richness. Finally, we showcase the realworld effectiveness of our approach, demonstrating an improved accuracy in solving the NavierStokes equations using a TF model with only a single parameter added to each encoding operation. Since TF models encompass conventional fixed frequency models, they may offer a sensible default choice for variational quantum machine learning.
Denoising Diffusion Probabilistic Models for HardwareImpaired Communications ; Generative AI has received significant attention among a spectrum of diverse industrial and academic domains, thanks to the magnificent results achieved from deep generative models such as generative pretrained transformers GPT and diffusion models. In this paper, we explore the applications of denoising diffusion probabilistic models DDPMs in wireless communication systems under practical assumptions such as hardware impairments HWI, lowSNR regime, and quantization error. Diffusion models are a new class of stateoftheart generative models that have already showcased notable success with some of the popular examples by OpenAI and Google Brain. The intuition behind DDPM is to decompose the data generation process over small denoising steps. Inspired by this, we propose using denoising diffusion modelbased receiver for a practical wireless communication scheme, while providing network resilience in lowSNR regimes, nonGaussian noise, different HWI levels, and quantization error. We evaluate the reconstruction performance of our scheme in terms of bit error rate BER and meansquared error MSE. Our results show that 30 and 20 improvement in BER could be achieved compared to deep neural network DNNbased receivers in AWGN and nonGaussian scenarios, respectively.
Learning Using Generated Privileged Information by TexttoImage Diffusion Models ; Learning Using Privileged Information is a particular type of knowledge distillation where the teacher model benefits from an additional data representation during training, called privileged information, improving the student model, which does not see the extra representation. However, privileged information is rarely available in practice. To this end, we propose a text classification framework that harnesses texttoimage diffusion models to generate artificial privileged information. The generated images and the original text samples are further used to train multimodal teacher models based on stateoftheart transformerbased architectures. Finally, the knowledge from multimodal teachers is distilled into a textbased unimodal student. Hence, by employing a generative model to produce synthetic data as privileged information, we guide the training of the student model. Our framework, called Learning Using Generated Privileged Information LUGPI, yields noticeable performance gains on four text classification data sets, demonstrating its potential in text classification without any additional cost during inference.
Long Text Generation via Adversarial Training with Leaked Information ; Automatically generating coherent and semantically meaningful text has many applications in machine translation, dialogue systems, image captioning, etc. Recently, by combining with policy gradient, Generative Adversarial Nets GAN that use a discriminative model to guide the training of the generative model as a reinforcement learning policy has shown promising results in text generation. However, the scalar guiding signal is only available after the entire text has been generated and lacks intermediate information about text structure during the generative process. As such, it limits its success when the length of the generated text samples is long more than 20 words. In this paper, we propose a new framework, called LeakGAN, to address the problem for long text generation. We allow the discriminative net to leak its own highlevel extracted features to the generative net to further help the guidance. The generator incorporates such informative signals into all generation steps through an additional Manager module, which takes the extracted features of current generated words and outputs a latent vector to guide the Worker module for nextword generation. Our extensive experiments on synthetic data and various realworld tasks with Turing test demonstrate that LeakGAN is highly effective in long text generation and also improves the performance in short text generation scenarios. More importantly, without any supervision, LeakGAN would be able to implicitly learn sentence structures only through the interaction between Manager and Worker.
Generative Adversarial Networks with DecoderEncoder Output Noise ; In recent years, research on image generation methods has been developing fast. The autoencoding variational Bayes method VAEs was proposed in 2013, which uses variational inference to learn a latent space from the image database and then generates images using the decoder. The generative adversarial networks GANs came out as a promising framework, which uses adversarial training to improve the generative ability of the generator. However, the generated pictures by GANs are generally blurry. The deep convolutional generative adversarial networks DCGANs were then proposed to leverage the quality of generated images. Since the input noise vectors are randomly sampled from a Gaussian distribution, the generator has to map from a whole normal distribution to the images. This makes DCGANs unable to reflect the inherent structure of the training data. In this paper, we propose a novel deep model, called generative adversarial networks with decoderencoder output noise DEGANs, which takes advantage of both the adversarial training and the variational Bayesain inference to improve the performance of image generation. DEGANs use a pretrained decoderencoder architecture to map the random Gaussian noise vectors to informative ones and pass them to the generator of the adversarial networks. Since the decoderencoder architecture is trained by the same images as the generators, the output vectors could carry the intrinsic distribution information of the original images. Moreover, the loss function of DEGANs is different from GANs and DCGANs. A hiddenspace loss function is added to the adversarial loss function to enhance the robustness of the model. Extensive empirical results show that DEGANs can accelerate the convergence of the adversarial training process and improve the quality of the generated images.
Aproximacion Discreta de la Relatividad General ; These Lecture notes give an introduction to Regge calculus as a discrete model of General Relativity.
Strict Genericity ; We show that an inner model of a classgeneric extension of L need not itself be such an extension. Our example is of the form LR, where R is a real belonging to a classgeneric extension of L and constructible from 0.
Quantum equivalence in PoissonLie Tduality ; We prove that, general smodels related by PoissonLie Tduality are quantum equivalent under oneloop renormalization group flow. We reveal general properties of the flows, we study the associated generalized coset models and provide explicit examples.