Datasets:
sswt
/
arxiv_sample_50K

Dataset card Data Studio Files
xet
Community
1
text
stringlengths
62
2.94k
ChatGPTsteered Editing Instructor for Customization of Abstractive Summarization ; Tailoring outputs of large language models, such as ChatGPT, to specific user needs remains a challenge despite their impressive generation quality. In this paper, we propose a triagent generation pipeline consisting of a generator, an instructor, and an editor to enhance the customization of generated outputs. The generator produces an initial output, the userspecific instructor generates editing instructions, and the editor generates a revised output aligned with user preferences. The inferenceonly large language model ChatGPT serves as both the generator and the editor, while a smaller model acts as the userspecific instructor to guide the generation process toward user needs. The instructor is trained using editorsteered reinforcement learning, leveraging feedback from the largescale editor model to optimize instruction generation. Experimental results on two abstractive summarization datasets demonstrate the effectiveness of our approach in generating outputs that better fulfill user expectations.
Retrieval Augmented Chest XRay Report Generation using OpenAI GPT models ; We propose Retrieval Augmented Generation RAG as an approach for automated radiology report writing that leverages multimodally aligned embeddings from a contrastively pretrained vision language model for retrieval of relevant candidate radiology text for an input radiology image and a general domain generative model like OpenAI textdavinci003, gpt3.5turbo and gpt4 for report generation using the relevant radiology text retrieved. This approach keeps hallucinated generations under check and provides capabilities to generate report content in the format we desire leveraging the instruction following capabilities of these generative models. Our approach achieves better clinical metrics with a BERTScore of 0.2865 Delta 25.88 and Semb score of 0.4026 Delta 6.31. Our approach can be broadly relevant for different clinical settings as it allows to augment the automated radiology report generation process with content relevant for that setting while also having the ability to inject user intents and requirements in the prompts as part of the report generation process to modulate the content and format of the generated reports as applicable for that clinical setting.
Generating Personalized Insulin Treatments Strategies with Deep Conditional Generative Time Series Models ; We propose a novel framework that combines deep generative time series models with decision theory for generating personalized treatment strategies. It leverages historical patient trajectory data to jointly learn the generation of realistic personalized treatment and future outcome trajectories through deep generative time series models. In particular, our framework enables the generation of novel multivariate treatment strategies tailored to the personalized patient history and trained for optimal expected future outcomes based on conditional expected utility maximization. We demonstrate our framework by generating personalized insulin treatment strategies and blood glucose predictions for hospitalized diabetes patients, showcasing the potential of our approach for generating improved personalized treatment strategies. Keywords deep generative model, probabilistic decision support, personalized treatment generation, insulin and blood glucose prediction
CausalGAN Learning Causal Implicit Generative Models with Adversarial Training ; We propose an adversarial training procedure for learning a causal implicit generative model for a given causal graph. We show that adversarial training can be used to learn a generative model with true observational and interventional distributions if the generator architecture is consistent with the given causal graph. We consider the application of generating faces based on given binary labels where the dependency structure between the labels is preserved with a causal graph. This problem can be seen as learning a causal implicit generative model for the image and labels. We devise a twostage procedure for this problem. First we train a causal implicit generative model over binary labels using a neural network consistent with a causal graph as the generator. We empirically show that WassersteinGAN can be used to output discrete labels. Later, we propose two new conditional GAN architectures, which we call CausalGAN and CausalBEGAN. We show that the optimal generator of the CausalGAN, given the labels, samples from the image distributions conditioned on these labels. The conditional GAN combined with a trained causal implicit generative model for the labels is then a causal implicit generative model over the labels and the generated image. We show that the proposed architectures can be used to sample from observational and interventional image distributions, even for interventions which do not naturally occur in the dataset.
MultiPLE A Scalable and Extensible Approach to Benchmarking Neural Code Generation ; Large language models have demonstrated the ability to generate both natural language and programming language text. Such models open up the possibility of multilanguage code generation could code generation models generalize knowledge from one language to another Although contemporary code generation models can generate semantically correct Python code, little is known about their abilities with other languages. We propose MultiPLE, a system for translating unit testdriven code generation benchmarks to new languages. We create the first massively multilingual code generation benchmark by using MultiPLE to translate two popular Python code generation benchmarks to 18 additional programming languages. We use MultiPLE to extend the HumanEval benchmark and MBPP benchmark to 18 languages that encompass a range of programming paradigms and popularity. Using these new parallel benchmarks, we evaluate the multilanguage performance of three stateoftheart code generation models Codex, CodeGen, and InCoder. We find that Codex matches or even exceeds its performance on Python for several other languages. The range of programming languages represented in MultiPLE allow us to explore the impact of language frequency and language features on model performance. Finally, the MultiPLE approach of compiling code generation benchmarks to new programming languages is both scalable and extensible, making it straightforward to evaluate new models, benchmarks, and languages.
Combinatorial Modelling and Learning with Prediction Markets ; Combining models in appropriate ways to achieve high performance is commonly seen in machine learning fields today. Although a large amount of combinatorial models have been created, little attention is drawn to the commons in different models and their connections. A general modelling technique is thus worth studying to understand model combination deeply and shed light on creating new models. Prediction markets show a promise of becoming such a generic, flexible combinatorial model. By reviewing on several popular combinatorial models and prediction market models, this paper aims to show how the market models can generalise different combinatorial stuctures and how they implement these popular combinatorial models in specific conditions. Besides, we will see among different market models, Storkey's emphMachine Learning Markets provide more fundamental, generic modelling mechanisms than the others, and it has a significant appeal for both theoretical study and application.
A ModuleSystem Discipline for ModelDriven Software Development ; Modeldriven development is a pragmatic approach to software development that embraces domainspecific languages DSLs, where models correspond to DSL programs. A distinguishing feature of modeldriven development is that clients of a model can select from an open set of alternative semantics of the model by applying different model transformation. However, in existing modeldriven frameworks, dependencies between models, model transformations, and generated code artifacts are either implicit or globally declared in build scripts, which impedes modular reasoning, separate compilation, and programmability in general. We propose the design of a new module system that incorporates models and model transformations as modules. A programmer can apply transformations in import statements, thus declaring a dependency on generated code artifacts. Our design enables modular reasoning and separate compilation by preventing hidden dependencies, and it supports mixing modeling artifacts with conventional code artifacts as well as higherorder transformations. We have formalized our design and the aforementioned properties and have validated it by an implementation and case studies that show that our module system successfully integrates modeldriven development into conventional programming languages.
GAMIN An Adversarial Approach to BlackBox Model Inversion ; Recent works have demonstrated that machine learning models are vulnerable to model inversion attacks, which lead to the exposure of sensitive information contained in their training dataset. While some model inversion attacks have been developed in the past in the blackbox attack setting, in which the adversary does not have direct access to the structure of the model, few of these have been conducted so far against complex models such as deep neural networks. In this paper, we introduce GAMIN for Generative Adversarial Model INversion, a new blackbox model inversion attack framework achieving significant results even against deep models such as convolutional neural networks at a reasonable computing cost. GAMIN is based on the continuous training of a surrogate model for the target model under attack and a generator whose objective is to generate inputs resembling those used to train the target model. The attack was validated against various neural networks used as image classifiers. In particular, when attacking models trained on the MNIST dataset, GAMIN is able to extract recognizable digits for up to 60 of labels produced by the target. Attacks against skin classification models trained on the pilot parliament dataset also demonstrated the capacity to extract recognizable features from the targets.
Automated Conversion of Axiomatic to Operational Models Theory and Practice ; A system may be modelled as an operational model which has explicit notions of state and transitions between states or an axiomatic model which is specified entirely as a set of invariants. Most formal methods techniques e.g., IC3, invariant synthesis, etc are designed for operational models and are largely inaccessible to axiomatic models. Furthermore, no prior method exists to automatically convert axiomatic models to operational ones, so operational equivalents to axiomatic models had to be manually created and proven equivalent. In this paper, we advance the stateoftheart in axiomatic to operational model conversion. We show that general axioms in the muspec axiomatic modelling framework cannot be translated to equivalent finitestate operational models. We also derive restrictions on the space of muspec axioms that enable the feasible generation of equivalent finitestate operational models for them. As for practical results, we develop a methodology for automatically translating muspec axioms to equivalent finitestate automatabased operational models. We demonstrate the efficacy of our method by using the models generated by our procedure to prove the correctness of ordering properties on three RTL designs.
The generic model of General Relativity ; We develop a generic spacetime model in General Relativity which can be used to build any gravitational model within General Relativity. The generic model uses two types of assumptions a Geometric assumptions additional to the inherent geometric identities of the Riemannian geometry of spacetime and b Assumptions defining a class of observers by means of their 4velocity ua which is a unit timelike vector field. The geometric assumptions as a rule concern symmetry assumptions the so called collineations. The latter introduces the 13 decomposition of tensor fields in spacetime. The 13 decomposition results in two major results. The 13 decomposition of ua;b defines the kinematic variables of the model expansion, rotation, shear and 4acceleration and defines the kinematics of the gravitational model. The 13 decomposition of the energy momentum tensor representing all gravitating matter introduces the dynamic variables of the model energy density, the isotropic pressure, the momentum transfer or heat flux vector and the traceless tensor of the anisotropic pressure as measured by the defined observers and define the dynamics of he model. The symmetries assumed by the model act as constraints on both the kinematical and the dynamical variables of the model. As a second further development of the generic model we assume that in addition to the 4velocity of the observers ua there exists a second universal vector field na in spacetime so that one has a so called double congruence ua,na which can be used to define the 112 decomposition of tensor fields. The 112 decomposition leads to an extended kinematics concerning both fields building the double congruence and to a finer dynamics involving more physical variables.
Domain Generalization using Pretrained Models without Finetuning ; Finetuning pretrained models is a common practice in domain generalization DG tasks. However, finetuning is usually computationally expensive due to the evergrowing size of pretrained models. More importantly, it may cause overfitting on source domain and compromise their generalization ability as shown in recent works. Generally, pretrained models possess some level of generalization ability and can achieve decent performance regarding specific domains and samples. However, the generalization performance of pretrained models could vary significantly over different test domains even samples, which raises challenges for us to best leverage pretrained models in DG tasks. In this paper, we propose a novel domain generalization paradigm to better leverage various pretrained models, named specialized ensemble learning for domain generalization SEDGE. It first trains a linear label space adapter upon fixed pretrained models, which transforms the outputs of the pretrained model to the label space of the target domain. Then, an ensemble network aware of model specialty is proposed to dynamically dispatch proper pretrained models to predict each test sample. Experimental studies on several benchmarks show that SEDGE achieves significant performance improvements comparing to strong baselines including stateoftheart method in DG tasks and reduces the trainable parameters by 99 and the training time by 99.5.
Generative AI for EndtoEnd Limit Order Book Modelling A TokenLevel Autoregressive Generative Model of Message Flow Using a Deep State Space Network ; Developing a generative model of realistic order flow in financial markets is a challenging open problem, with numerous applications for market participants. Addressing this, we propose the first endtoend autoregressive generative model that generates tokenized limit order book LOB messages. These messages are interpreted by a JaxLOB simulator, which updates the LOB state. To handle long sequences efficiently, the model employs simplified structured statespace layers to process sequences of order book states and tokenized messages. Using LOBSTER data of NASDAQ equity LOBs, we develop a custom tokenizer for message data, converting groups of successive digits to tokens, similar to tokenization in large language models. Outofsample results show promising performance in approximating the data distribution, as evidenced by low model perplexity. Furthermore, the midprice returns calculated from the generated order flow exhibit a significant correlation with the data, indicating impressive conditional forecast performance. Due to the granularity of generated data, and the accuracy of the model, it offers new application areas for future work beyond forecasting, e.g. acting as a world model in highfrequency financial reinforcement learning applications. Overall, our results invite the use and extension of the model in the direction of autoregressive large financial models for the generation of highfrequency financial data and we commit to opensourcing our code to facilitate future research.
A unified view of generative models for networks models, methods, opportunities, and challenges ; Research on probabilistic models of networks now spans a wide variety of fields, including physics, sociology, biology, statistics, and machine learning. These efforts have produced a diverse ecology of models and methods. Despite this diversity, many of these models share a common underlying structure pairwise interactions edges are generated with probability conditional on latent vertex attributes. Differences between models generally stem from different philosophical choices about how to learn from data or different empiricallymotivated goals. The highly interdisciplinary nature of work on these generative models, however, has inhibited the development of a unified view of their similarities and differences. For instance, novel theoretical models and optimization techniques developed in machine learning are largely unknown within the social and biological sciences, which have instead emphasized model interpretability. Here, we describe a unified view of generative models for networks that draws together many of these disparate threads and highlights the fundamental similarities and differences that span these fields. We then describe a number of opportunities and challenges for future work that are revealed by this view.
An Attentional Neural Conversation Model with Improved Specificity ; In this paper we propose a neural conversation model for conducting dialogues. We demonstrate the use of this model to generate help desk responses, where users are asking questions about PC applications. Our model is distinguished by two characteristics. First, it models intention across turns with a recurrent network, and incorporates an attention model that is conditioned on the representation of intention. Secondly, it avoids generating nonspecific responses by incorporating an IDF term in the objective function. The model is evaluated both as a pure generation model in which a helpdesk response is generated from scratch, and as a retrieval model with performance measured using recall rates of the correct response. Experimental results indicate that the model outperforms previously proposed neural conversation architectures, and that using specificity in the objective function significantly improves performances for both generation and retrieval.
Connectivity and Structure in Large Networks ; Large reallife complex networks are often modeled by various random graph constructions and hundreds of further references therein. In many cases it is not at all clear how the modeling strength of differently generated random graph model classes relate to each other. We would like to systematically investigate such issues. Our approach was originally motivated to capture properties of the random network topology of wireless communication networks. We started some investigations, but here we elevate it to a more general level that makes it possible to compare the strength of different classes of random network models. Specially, we introduce various classes of random graph models that are significantly more general than the ones that are usually treated in the literature, and show relationships among them. One of our main results is that no random graph model can fall in the following three classes at the same time 1 random graph models with bounded expected degrees; 2 random graph models that are asymptotically almost connected; 3 an abstracted version of geometric random graph models with two mild restrictions that we call locality and name invariance. In other words, in a mildly restricted, but still very general, class of generalized geometricstyle models the requirements of bounded expected degrees and asymptotic almost connectivity are incompatible.
Marginally Interpretable Generalized Linear Mixed Models ; Two popular approaches for relating correlated measurements of a nonGaussian response variable to a set of predictors are to fit a marginal model using generalized estimating equations and to fit a generalized linear mixed model by introducing latent random variables. The first approach is effective for parameter estimation, but leaves one without a formal model for the data with which to assess quality of fit or make predictions for future observations. The second approach overcomes the deficiencies of the first, but leads to parameter estimates that must be interpreted conditional on the latent variables. Further complicating matters, obtaining marginal summaries from a generalized linear mixed model often requires evaluation of an analytically intractable integral or use of attenuation factors that are not exact. We define a class of marginally interpretable generalized linear mixed models that lead to parameter estimates with a marginal interpretation while maintaining the desirable statistical properties of a conditionallyspecified model. We discuss the form of these models under various common link functions and also address computational issues associated with these models. For logistic mixed effects models, we introduce an accurate and efficient method for evaluating the logisticnormal integral.
Learning Dynamic Generator Model by Alternating BackPropagation Through Time ; This paper studies the dynamic generator model for spatialtemporal processes such as dynamic textures and action sequences in video data. In this model, each time frame of the video sequence is generated by a generator model, which is a nonlinear transformation of a latent state vector, where the nonlinear transformation is parametrized by a topdown neural network. The sequence of latent state vectors follows a nonlinear autoregressive model, where the state vector of the next frame is a nonlinear transformation of the state vector of the current frame as well as an independent noise vector that provides randomness in the transition. The nonlinear transformation of this transition model can be parametrized by a feedforward neural network. We show that this model can be learned by an alternating backpropagation through time algorithm that iteratively samples the noise vectors and updates the parameters in the transition model and the generator model. We show that our training method can learn realistic models for dynamic textures and action patterns.
VideoFlow A Conditional FlowBased Model for Stochastic Video Generation ; Generative models that can model and predict sequences of future events can, in principle, learn to capture complex realworld phenomena, such as physical interactions. However, a central challenge in video prediction is that the future is highly uncertain a sequence of past observations of events can imply many possible futures. Although a number of recent works have studied probabilistic models that can represent uncertain futures, such models are either extremely expensive computationally as in the case of pixellevel autoregressive models, or do not directly optimize the likelihood of the data. To our knowledge, our work is the first to propose multiframe video prediction with normalizing flows, which allows for direct optimization of the data likelihood, and produces highquality stochastic predictions. We describe an approach for modeling the latent space dynamics, and demonstrate that flowbased generative models offer a viable and competitive approach to generative modelling of video.
Zeroshot TexttoSQL Learning with Auxiliary Task ; Recent years have seen great success in the use of neural seq2seq models on the texttoSQL task. However, little work has paid attention to how these models generalize to realistic unseen data, which naturally raises a question does this impressive performance signify a perfect generalization model, or are there still some limitations In this paper, we first diagnose the bottleneck of texttoSQL task by providing a new testbed, in which we observe that existing models present poor generalization ability on rarelyseen data. The above analysis encourages us to design a simple but effective auxiliary task, which serves as a supportive model as well as a regularization term to the generation task to increase the models generalization. Experimentally, We evaluate our models on a large texttoSQL dataset WikiSQL. Compared to a strong baseline coarsetofine model, our models improve over the baseline by more than 3 absolute in accuracy on the whole dataset. More interestingly, on a zeroshot subset test of WikiSQL, our models achieve 5 absolute accuracy gain over the baseline, clearly demonstrating its superior generalizability.
Powering Hidden Markov Model by Neural Network based Generative Models ; Hidden Markov model HMM has been successfully used for sequential data modeling problems. In this work, we propose to power the modeling capacity of HMM by bringing in neural network based generative models. The proposed model is termed as GenHMM. In the proposed GenHMM, each HMM hidden state is associated with a neural network based generative model that has tractability of exact likelihood and provides efficient likelihood computation. A generative model in GenHMM consists of mixture of generators that are realized by flow models. A learning algorithm for GenHMM is proposed in expectationmaximization framework. The convergence of the learning GenHMM is analyzed. We demonstrate the efficiency of GenHMM by classification tasks on practical sequential data. Code available at httpsgithub.comFirstHandScientistgenhmm.
Empathetic BERT2BERT Conversational Model Learning Arabic Language Generation with Little Data ; Enabling empathetic behavior in Arabic dialogue agents is an important aspect of building humanlike conversational models. While Arabic Natural Language Processing has seen significant advances in Natural Language Understanding NLU with language models such as AraBERT, Natural Language Generation NLG remains a challenge. The shortcomings of NLG encoderdecoder models are primarily due to the lack of Arabic datasets suitable to train NLG models such as conversational agents. To overcome this issue, we propose a transformerbased encoderdecoder initialized with AraBERT parameters. By initializing the weights of the encoder and decoder with AraBERT pretrained weights, our model was able to leverage knowledge transfer and boost performance in response generation. To enable empathy in our conversational model, we train it using the ArabicEmpatheticDialogues dataset and achieve high performance in empathetic response generation. Specifically, our model achieved a low perplexity value of 17.0 and an increase in 5 BLEU points compared to the previous stateoftheart model. Also, our proposed model was rated highly by 85 human evaluators, validating its high capability in exhibiting empathy while generating relevant and fluent responses in opendomain settings.
GENOME A GENeric methodology for Ontological Modelling of Epics ; Ontological knowledge modelling of epics, though being an established research arena backed by concrete multilingual and multicultural works, still suffer from two key shortcomings. Firstly, all epic ontological models developed till date have been designed following adhoc methodologies, most often, combining existing general purpose ontology development methodologies. Secondly, none of the adhoc methodologies consider the potential reuse of existing epic ontological models for enrichment, if available. The paper presents, as a unified solution to the above shortcomings, the design and development of GENOME the first dedicated methodology for iterative ontological modelling of epics, potentially extensible to works in different research arenas of digital humanities in general. GENOME is grounded in transdisciplinary foundations of canonical norms for epics, knowledge modelling best practices, application satisfiability norms and cognitive generative questions. It is also the first methodology in epic modelling but also in general to be flexible enough to integrate, in practice, the options of knowledge modelling via reuse or from scratch. The feasibility of GENOME is validated via a first brief implementation of ontological modelling of the Indian epic Mahabharata by reusing an existing ontology. The preliminary results are promising, with the GENOMEproduced model being both ontologically thorough and performancewise competent
Learning to Model Editing Processes ; Most existing sequence generation models produce outputs in one pass, usually lefttoright. However, this is in contrast with a more natural approach that humans use in generating content; iterative refinement and editing. Recent work has introduced editbased models for various tasks such as neural machine translation and text style transfer, but these generally model a single edit step. In this work, we propose modeling editing processes, modeling the whole process of iteratively generating sequences. We form a conceptual framework to describe the likelihood of multistep edits, and describe neural models that can learn a generative model of sequences based on these multistep edits. We introduce baseline results and metrics on this task, finding that modeling editing processes improves performance on a variety of axes on both our proposed task and related downstream tasks compared to previous singlestep models of edits.
Interactive Text Generation ; Users interact with text, image, code, or other editors on a daily basis. However, machine learning models are rarely trained in the settings that reflect the interactivity between users and their editor. This is understandable as training AI models with real users is not only slow and costly, but what these models learn may be specific to user interface design choices. Unfortunately, this means most of the research on text, code, and image generation has focused on noninteractive settings, whereby the model is expected to get everything right without accounting for any input from a user who may be willing to help. We introduce a new Interactive Text Generation task that allows training generation models interactively without the costs of involving real users, by using user simulators that provide edits that guide the model towards a given target text. We train our interactive models using Imitation Learning, and our experiments against competitive noninteractive generation models show that models trained interactively are superior to their noninteractive counterparts, even when all models are given the same budget of user inputs or edits.
A Probabilistic Fluctuation based Membership Inference Attack for Diffusion Models ; Membership Inference Attack MIA identifies whether a record exists in a machine learning model's training set by querying the model. MIAs on the classic classification models have been wellstudied, and recent works have started to explore how to transplant MIA onto generative models. Our investigation indicates that existing MIAs designed for generative models mainly depend on the overfitting in target models. However, overfitting can be avoided by employing various regularization techniques, whereas existing MIAs demonstrate poor performance in practice. Unlike overfitting, memorization is essential for deep learning models to attain optimal performance, making it a more prevalent phenomenon. Memorization in generative models leads to an increasing trend in the probability distribution of generating records around the member record. Therefore, we propose a Probabilistic Fluctuation Assessing Membership Inference Attack PFAMI, a blackbox MIA that infers memberships by detecting these trends via analyzing the overall probabilistic fluctuations around given records. We conduct extensive experiments across multiple generative models and datasets, which demonstrate PFAMI can improve the attack success rate ASR by about 27.9 when compared with the best baseline.
Generative Quantum Machine Learning ; The goal of generative machine learning is to model the probability distribution underlying a given data set. This probability distribution helps to characterize the generation process of the data samples. While classical generative machine learning is solely based on classical resources, generative quantum machine learning can also employ quantum resources such as parameterized quantum channels and quantum operators to learn and sample from the probability model of interest. Applications of generative quantum models are multifaceted. The trained model can generate new samples that are compatible with the given data and extend the data set. Additionally, learning a model for the generation process of a data set may provide interesting information about the corresponding properties. With the help of quantum resources, the respective generative models have access to functions that are difficult to evaluate with a classical computer and may improve the performance or lead to new insights. Furthermore, generative quantum machine learning can be applied to efficient, approximate loading of classical data into a quantum state which may help to avoid potentially exponentially expensive, exact quantum data loading. The aim of this doctoral thesis is to develop new generative quantum machine learning algorithms, demonstrate their feasibility, and analyze their performance. Additionally, we outline their potential application to efficient, approximate quantum data loading. More specifically, we introduce a quantum generative adversarial network and a quantum Boltzmann machine implementation, both of which can be realized with parameterized quantum circuits. These algorithms are compatible with firstgeneration quantum hardware and, thus, enable us to study proof of concept implementations not only with numerical quantum simulations but also real quantum hardware available today.
ConvGeN Convex space learning improves deepgenerative oversampling for tabular imbalanced classification on smaller datasets ; Data is commonly stored in tabular format. Several fields of research are prone to small imbalanced tabular data. Supervised Machine Learning on such data is often difficult due to class imbalance. Synthetic data generation, i.e., oversampling, is a common remedy used to improve classifier performance. Stateoftheart linear interpolation approaches, such as LoRAS and ProWRAS can be used to generate synthetic samples from the convex space of the minority class to improve classifier performance in such cases. Deep generative networks are common deep learning approaches for synthetic sample generation, widely used for synthetic image generation. However, their scope on synthetic tabular data generation in the context of imbalanced classification is not adequately explored. In this article, we show that existing deep generative models perform poorly compared to linear interpolation based approaches for imbalanced classification problems on smaller tabular datasets. To overcome this, we propose a deep generative model, ConvGeN that combines the idea of convex space learning with deep generative models. ConvGeN learns the coefficients for the convex combinations of the minority class samples, such that the synthetic data is distinct enough from the majority class. Our benchmarking experiments demonstrate that our proposed model ConvGeN improves imbalanced classification on such small datasets, as compared to existing deep generative models, while being atpar with the existing linear interpolation approaches. Moreover, we discuss how our model can be used for synthetic tabular data generation in general, even outside the scope of data imbalance and thus, improves the overall applicability of convex space learning.
On the Reliability and Explainability of Automated Code Generation Approaches ; Automatic code generation, the task of generating new code snippets from existing code or comments, has long been of interest. Numerous code generation models have been proposed and proven on different benchmark datasets. However, little is known about whether this objective has been achieved and why code generation models effectively transform code sequences automatically. In other words, can we totally trust these automated code generation models Consequently, there is a pressing need to understand the inner logic of code generation models and to investigate their replicability, reliability, and explainability. To bridge these research gaps, we conduct a thorough empirical study of five code generation models on four representative code generation datasets to assess the limits and capabilities of automatic code generation approaches. We further employ advanced explainable AI approaches to highlight the tokens that significantly contribute to the code generation. Experiments demonstrate that we successfully replicate stateoftheart code generation approaches. We discover that stateoftheart approaches suffer from severe data duplication and input insensitivity, which are subtle issues with significant implications. Our explainability analysis reveals that, in various experimental scenarios, code generation models can recognize code grammar and structural information, but can not capture key tokens that need to be updated. Our results draw several lessons and guidelines for future work in this area.
Stability in Generalized Modified Gravity ; The stability issue of a large class of modified gravitational models is discussed with particular emphasis to de Sitter solutions. Three approaches are briefly presented and the generalization to more general cases is mentioned.
Generalized Network Psychometrics Combining Network and Latent Variable Models ; We introduce the network model as a formal psychometric model, conceptualizing the covariance between psychometric indicators as resulting from pairwise interactions between observable variables in a network structure. This contrasts with standard psychometric models, in which the covariance between test items arises from the influence of one or more common latent variables. Here, we present two generalizations of the network model that encompass latent variable structures, establishing network modeling as parts of the more general framework of Structural Equation Modeling SEM. In the first generalization, we model the covariance structure of latent variables as a network. We term this framework Latent Network Modeling LNM and show that, with LNM, a unique structure of conditional independence relationships between latent variables can be obtained in an explorative manner. In the second generalization, the residual variancecovariance structure of indicators is modeled as a network. We term this generalization Residual Network Modeling RNM and show that, within this framework, identifiable models can be obtained in which local independence is structurally violated. These generalizations allow for a general modeling framework that can be used to fit, and compare, SEM models, network models, and the RNM and LNM generalizations. This methodology has been implemented in the freetouse software package lvnet, which contains confirmatory model testing as well as two exploratory search algorithms stepwise search algorithms for lowdimensional datasets and penalized maximum likelihood estimation for larger datasets. We show in simulation studies that these search algorithms performs adequately in identifying the structure of the relevant residual or latent networks. We further demonstrate the utility of these generalizations in an empirical example on a personality inventory dataset.
ExactlySolvable Models Derived from a Generalized Gaudin Algebra ; We introduce a generalized Gaudin Lie algebra and a complete set of mutually commuting quantum invariants allowing the derivation of several families of exactly solvable Hamiltonians. Different Hamiltonians correspond to different representations of the generators of the algebra. The derived exactlysolvable generalized Gaudin models include the BardeenCooperSchrieffer, SuhlMatthiasWalker, the LipkinMeshkovGlick, generalized Dicke, the Nuclear Interacting Boson Model, a new exactlysolvable Kondolike impurity model, and many more that have not been exploited in the physics literature yet.
On the Use of Cellular Automata in Symmetric Cryptography ; In this work, pseudorandom sequence generators based on finite fields have been analyzed from the point of view of their cryptographic application. In fact, a class of nonlinear sequence generators has been modelled in terms of linear cellular automata. The algorithm that converts the given generator into a linear model based on automata is very simple and is based on the concatenation of a basic structure. Once the generator has been linearized, a cryptanalytic attack that exploits the weaknesses of such a model has been developed. Linear cellular structures easily model sequence generators with application in stream cipher cryptography.
Simple model for quantum general relativity from loop quantum gravity ; New progress in loop gravity has lead to a simple model of generalcovariant quantum field theory'. I sum up the definition of the model in selfcontained form, in terms accessible to those outside the subfield. I emphasize its formulation as a generalized topological quantum field theory with an infinite number of degrees of freedom, and its relation to lattice theory. I list the indications supporting the conjecture that the model is related to general relativity and UV finite.
Thermodynamics in KaluzaKlein Universe ; This paper is devoted to check the validity of laws of thermodynamics for KaluzaKlein universe in the state of thermal equilibrium, composed of dark matter and dark energy. The generalized holographic dark energy and generalized Ricci dark energy models are considered here. It is proved that the first and generalized second law of thermodynamics are valid on the apparent horizon for both of these models. Further, we take a horizon of radius L with modified holographic or Ricci dark energy. We conclude that these models do not obey the first and generalized second law of thermodynamics on the horizon of fixed radius L for a specific range of model parameters.
Morphological Inflection Generation Using Character Sequence to Sequence Learning ; Morphological inflection generation is the task of generating the inflected form of a given lemma corresponding to a particular linguistic transformation. We model the problem of inflection generation as a character sequence to sequence learning problem and present a variant of the neural encoderdecoder model for solving it. Our model is language independent and can be trained in both supervised and semisupervised settings. We evaluate our system on seven datasets of morphologically rich languages and achieve either better or comparable results to existing stateoftheart models of inflection generation.
Nonstandard cohomology for equivariant sheaves The role of generic models ; We generalize the Generic Model Theorem for equivariant presheaves of structures; extending the results of Macintyre and Caicedo. We also introduce a new class of generic cohomologies and show how, for some examples, they simplify to non standard cohomologies. Key words and phrases Generic Model Theorem, Equivariant Structures, Equivariant Cohomology. Primary fields Model Theory. Equivariant sheaf cohomology.
Generalized model for anisotropic compact stars ; In the present investigation an exact generalized model for anisotropic compact stars of embedding class one is sought for under general relativistic background. The generic solutions are verified by exploring different physical aspects, viz. energy conditions, massradius relation, stability of the models, in connection to their validity. It is observed that the model present here for compact stars is compatible with all these physical tests and thus physically acceptable as far as the compact star candidates RXJ185637, SAXJ1808.43658SS1 and SAXJ1808.43658SS2 are concerned.
Implicit Modeling A Generalization of Discriminative and Generative Approaches ; We propose a new modeling approach that is a generalization of generative and discriminative models. The core idea is to use an implicit parameterization of a joint probability distribution by specifying only the conditional distributions. The proposed scheme combines the advantages of both worlds it can use powerful complex discriminative models as its parts, having at the same time better generalization capabilities. We thoroughly evaluate the proposed method for a simple classification task with artificial data and illustrate its advantages for realword scenarios on a semantic image segmentation problem.
Generative Code Modeling with Graphs ; Generative models for source code are an interesting structured prediction problem, requiring to reason about both hard syntactic and semantic constraints as well as about natural, likely programs. We present a novel model for this problem that uses a graph to represent the intermediate state of the generated output. The generative procedure interleaves grammardriven expansion steps with graph augmentation and neural message passing steps. An experimental evaluation shows that our new model can generate semantically meaningful expressions, outperforming a range of strong baselines.
NIPS 2016 Tutorial Generative Adversarial Networks ; This report summarizes the tutorial presented by the author at NIPS 2016 on generative adversarial networks GANs. The tutorial describes 1 Why generative modeling is a topic worth studying, 2 how generative models work, and how GANs compare to other generative models, 3 the details of how GANs work, 4 research frontiers in GANs, and 5 stateoftheart image models that combine GANs with other methods. Finally, the tutorial contains three exercises for readers to complete, and the solutions to these exercises.
Human Motion Modeling using DVGANs ; We present a novel generative model for human motion modeling using Generative Adversarial Networks GANs. We formulate the GAN discriminator using dense validation at each timescale and perturb the discriminator input to make it translation invariant. Our model is capable of motion generation and completion. We show through our evaluations the resiliency to noise, generalization over actions, and generation of long diverse sequences. We evaluate our approach on Human 3.6M and CMU motion capture datasets using inception scores.
Introducing the Generalized GNmodel for Nonlinear Interference Generation including spacefrequency variations of lossgain ; We develop and present a generalization of the GNmodel the generalized Gaussian noise GGN model to enabling a fair application of GNmodel to predict generation of nonlinear interference when loss parameters relevantly vary with frequency andor distributed amplification applies selectively to portions of the exploited spectrum andor stimulatedRamanscatteringinduced crosstalk is relevant.
Locally optimal designs for generalized linear models within the family of Kiefer kcriteria ; Locally optimal designs for generalized linear models are derived at certain values of the regression parameters. In the present paper a general setup of the generalized linear model is considered. Analytic solutions for optimal designs are developed under Kiefer Phikcriteria highlighting the D and Aoptimal designs. By means of The General Equivalence Theorem necessary and sufficient conditions in term of intensity values are obtained to characterize the locally optimal designs. In this context, linear predictors are assumed constituting first order models with and without intercept on appropriate experimental regions.
Provable Lipschitz Certification for Generative Models ; We present a scalable technique for upper bounding the Lipschitz constant of generative models. We relate this quantity to the maximal norm over the set of attainable vectorJacobian products of a given generative model. We approximate this set by layerwise convex approximations using zonotopes. Our approach generalizes and improves upon prior work using zonotope transformers and we extend to Lipschitz estimation of neural networks with large output dimension. This provides efficient and tight bounds on small networks and can scale to generative models on VAE and DCGAN architectures.
Topcolorlike dynamics and new matter generations ; We explore a scenarios where topcolorlike dynamics operates in the presence of fourth generation matter fields. Using the Minimal Walking Technicolor as a concrete basis for model building, we construct explicit models and confront them with phenomenology. We show that if a new QCD generations exist, both the topbottom mass splitting as well as the splitting between bottom quark mass and the masses of the fourth generation quarks can be naturally explained within topcolorlike dynamics. On the other hand, the much studied Minimal Walking Technicolor model where only a fourth generation of leptons arise, also leads to a viable model.
DPAGE Diverse Paraphrase Generation ; In this paper, we investigate the diversity aspect of paraphrase generation. Prior deep learning models employ either decoding methods or add random input noise for varying outputs. We propose a simple method Diverse Paraphrase Generation DPAGE, which extends neural machine translation NMT models to support the generation of diverse paraphrases with implicit rewriting patterns. Our experimental results on two realworld benchmark datasets demonstrate that our model generates at least one order of magnitude more diverse outputs than the baselines in terms of a new evaluation metric Jeffrey's Divergence. We have also conducted extensive experiments to understand various properties of our model with a focus on diversity.
Improving Neural Question Generation using World Knowledge ; In this paper, we propose a method for incorporating world knowledge linked entities and finegrained entity types into a neural question generation model. This world knowledge helps to encode additional information related to the entities present in the passage required to generate humanlike questions. We evaluate our models on both SQuAD and MS MARCO to demonstrate the usefulness of the world knowledge features. The proposed world knowledge enriched question generation model is able to outperform the vanilla neural question generation model by 1.37 and 1.59 absolute BLEU 4 score on SQuAD and MS MARCO test dataset respectively.
TMLab Generative Enhanced Model GEM for adversarial attacks ; We present our Generative Enhanced Model GEM that we used to create samples awarded the first prize on the FEVER 2.0 Breakers Task. GEM is the extended language model developed upon GPT2 architecture. The addition of novel target vocabulary input to the already existing context input enabled controlled text generation. The training procedure resulted in creating a model that inherited the knowledge of pretrained GPT2, and therefore was ready to generate naturallike English sentences in the task domain with some additional control. As a result, GEM generated malicious claims that mixed facts from various articles, so it became difficult to classify their truthfulness.
Label Dependent Deep Variational Paraphrase Generation ; Generating paraphrases that are lexically similar but semantically different is a challenging task. Paraphrases of this form can be used to augment data sets for various NLP tasks such as machine reading comprehension and question answering with nontrivial negative examples. In this article, we propose a deep variational model to generate paraphrases conditioned on a label that specifies whether the paraphrases are semantically related or not. We also present new training recipes and KL regularization techniques that improve the performance of variational paraphrasing models. Our proposed model demonstrates promising results in enhancing the generative power of the model by employing labeldependent generation on paraphrasing datasets.
Getting Topology and Point Cloud Generation to Mesh ; In this work, we explore the idea that effective generative models for point clouds under the autoencoding framework must acknowledge the relationship between a continuous surface, a discretized mesh, and a set of points sampled from the surface. This view motivates a generative model that works by progressively deforming a uniform sphere until it approximates the goal point cloud. We review the underlying concepts leading to this conclusion from computer graphics and topology in differential geometry, and model the generation process as deformation via deep neural network parameterization. Finally, we show that this view of the problem produces a model that can generate quality meshes efficiently.
Towards a Fast and Accurate Model of Intercontact Times for Epidemic Routing ; We present an accurate userencounter trace generator based on analytical models. Our method generates traces of intercontact times faster than models that explicitly generate mobility traces. We use this trace generator to study the characteristics of pairwise intercontacttime distributions and visualize, using simulations, how they combine to form the aggregate intercontacttime distribution. Finally, we apply our tracegeneration model to the epidemic routing protocol.
Hierarchical Video Generation for Complex Data ; Videos can often be created by first outlining a global description of the scene and then adding local details. Inspired by this we propose a hierarchical model for video generation which follows a coarse to fine approach. First our model generates a low resolution video, establishing the global scene structure, that is then refined by subsequent levels in the hierarchy. We train each level in our hierarchy sequentially on partial views of the videos. This reduces the computational complexity of our generative model, which scales to highresolution videos beyond a few frames. We validate our approach on Kinetics600 and BDD100K, for which we train a three level model capable of generating 256x256 videos with 48 frames.
A Unified Framework for Pun Generation with Humor Principles ; We propose a unified framework to generate both homophonic and homographic puns to resolve the splitup in existing works. Specifically, we incorporate three linguistic attributes of puns to the language models ambiguity, distinctiveness, and surprise. Our framework consists of three parts 1 a context wordsphrases selector to promote the aforementioned attributes, 2 a generation model trained on nonpun sentences to incorporate the context wordsphrases into the generation output, and 3 a label predictor that learns the structure of puns which is used to steer the generation model at inference time. Evaluation results on both pun types demonstrate the efficacy of our model over strong baselines.
LayoutDiffuse Adapting Foundational Diffusion Models for LayouttoImage Generation ; Layouttoimage generation refers to the task of synthesizing photorealistic images based on semantic layouts. In this paper, we propose LayoutDiffuse that adapts a foundational diffusion model pretrained on largescale image or textimage datasets for layouttoimage generation. By adopting a novel neural adaptor based on layout attention and taskaware prompts, our method trains efficiently, generates images with both high perceptual quality and layout alignment, and needs less data. Experiments on three datasets show that our method significantly outperforms other 10 generative models based on GANs, VQVAE, and diffusion models.
DiffECG A Generalized Probabilistic Diffusion Model for ECG Signals Synthesis ; In recent years, deep generative models have gained attention as a promising data augmentation solution for heart disease detection using deep learning approaches applied to ECG signals. In this paper, we introduce a novel approach based on denoising diffusion probabilistic models for ECG synthesis that covers three scenarios heartbeat generation, partial signal completion, and full heartbeat forecasting. Our approach represents the first generalized conditional approach for ECG synthesis, and our experimental results demonstrate its effectiveness for various ECGrelated tasks. Moreover, we show that our approach outperforms other stateoftheart ECG generative models and can enhance the performance of stateoftheart classifiers.
Improving Domain Generalization for Sound Classification with Sparse FrequencyRegularized Transformer ; Sound classification models' performance suffers from generalizing on outofdistribution OOD data. Numerous methods have been proposed to help the model generalize. However, most either introduce inference overheads or focus on longlasting CNNvariants, while Transformers has been proven to outperform CNNs on numerous natural language processing and computer vision tasks. We propose FRITO, an effective regularization technique on Transformer's selfattention, to improve the model's generalization ability by limiting each sequence position's attention receptive field along the frequency dimension on the spectrogram. Experiments show that our method helps Transformer models achieve SOTA generalization performance on TAU 2020 and Nsynth datasets while saving 20 inference time.
Stochastic Feature Mapping for PACBayes Classification ; Probabilistic generative modeling of data distributions can potentially exploit hidden information which is useful for discriminative classification. This observation has motivated the development of approaches that couple generative and discriminative models for classification. In this paper, we propose a new approach to couple generative and discriminative models in an unified framework based on PACBayes risk theory. We first derive the modelparameterindependent stochastic feature mapping from a practical MAP classifier operating on generative models. Then we construct a linear stochastic classifier equipped with the feature mapping, and derive the explicit PACBayes risk bounds for such classifier for both supervised and semisupervised learning. Minimizing the risk bound, using an EMlike iterative procedure, results in a new posterior over hidden variables Estep and the update rules of model parameters Mstep. The derivation of the posterior is always feasible due to the way of equipping feature mapping and the explicit form of bounding risk. The derived posterior allows the tuning of generative models and subsequently the feature mappings for better classification. The derived update rules of the model parameters are same to those of the uncoupled models as the feature mapping is modelparameterindependent. Our experiments show that the coupling between data modeling generative model and the discriminative classifier via a stochastic feature mapping in this framework leads to a general classification tool with stateoftheart performance.
Defuse Harnessing Unrestricted Adversarial Examples for Debugging Models Beyond Test Accuracy ; We typically compute aggregate statistics on heldout test data to assess the generalization of machine learning models. However, statistics on test data often overstate model generalization, and thus, the performance of deployed machine learning models can be variable and untrustworthy. Motivated by these concerns, we develop methods to automatically discover and correct model errors beyond those available in the data. We propose Defuse, a method that generates novel model misclassifications, categorizes these errors into highlevel model bugs, and efficiently labels and finetunes on the errors to correct them. To generate misclassified data, we propose an algorithm inspired by adversarial machine learning techniques that uses a generative model to find naturally occurring instances misclassified by a model. Further, we observe that the generative models have regions in their latent space with higher concentrations of misclassifications. We call these regions misclassification regions and find they have several useful properties. Each region contains a specific type of model bug; for instance, a misclassification region for an MNIST classifier contains a style of skinny 6 that the model mistakes as a 1. We can also assign a single label to each region, facilitating lowcost labeling. We propose a method to learn the misclassification regions and use this insight to both categorize errors and correct them. In practice, Defuse finds and corrects novel errors in classifiers. For example, Defuse shows that a highperformance traffic sign classifier mistakes certain 50kmh signs as 80kmh. Defuse corrects the error after finetuning while maintaining generalization on the test set.
Shortterm forecasting of solar irradiance without local telemetry a generalized model using satellite data ; Due to the increasing integration of solar power into the electrical grid, forecasting shortterm solar irradiance has become key for many applications, e.g.operational planning, power purchases, reserve activation, etc. In this context, as solar generators are geographically dispersed and ground measurements are not always easy to obtain, it is very important to have general models that can predict solar irradiance without the need of local data. In this paper, a model that can perform shortterm forecasting of solar irradiance in any general location without the need of ground measurements is proposed. To do so, the model considers satellitebased measurements and weatherbased forecasts, and employs a deep neural network structure that is able to generalize across locations; particularly, the network is trained only using a small subset of sites where ground data is available, and the model is able to generalize to a much larger number of locations where ground data does not exist. As a case study, 25 locations in The Netherlands are considered and the proposed model is compared against four local models that are individually trained for each location using ground measurements. Despite the general nature of the model, it is shown show that the proposed model is equal or better than the local models when comparing the average performance across all the locations and prediction horizons, the proposed model obtains a 31.31 rRMSE relative root mean square error while the best local model achieves a 32.01 rRMSE.
PETGEN Personalized Text Generation Attack on Deep Sequence Embeddingbased Classification Models ; What should a malicious user write next to fool a detection model Identifying malicious users is critical to ensure the safety and integrity of internet platforms. Several deep learningbased detection models have been created. However, malicious users can evade deep detection models by manipulating their behavior, rendering these models of little use. The vulnerability of such deep detection models against adversarial attacks is unknown. Here we create a novel adversarial attack model against deep user sequence embedding based classification models, which use the sequence of user posts to generate user embeddings and detect malicious users. In the attack, the adversary generates a new post to fool the classifier. We propose a novel endtoend Personalized Text Generation Attack model, called PETGEN, that simultaneously reduces the efficacy of the detection model and generates posts that have several key desirable properties. Specifically, PETGEN generates posts that are personalized to the user's writing style, have knowledge about a given target context, are aware of the user's historical posts on the target context, and encapsulate the user's recent topical interests. We conduct extensive experiments on two realworld datasets Yelp and Wikipedia, both with groundtruth of malicious users to show that PETGEN significantly reduces the performance of popular deep user sequence embeddingbased classification models. PETGEN outperforms five attack baselines in terms of text quality and attack efficacy in both whitebox and blackbox classifier settings. Overall, this work paves the path towards the next generation of adversaryaware sequence classification models.
A Survey on Generative Diffusion Model ; Deep generative models are a prominent approach for data generation, and have been used to produce high quality samples in various domains. Diffusion models, an emerging class of deep generative models, have attracted considerable attention owing to their exceptional generative quality. Despite this, they have certain limitations, including a timeconsuming iterative generation process and confinement to highdimensional Euclidean space. This survey presents a plethora of advanced techniques aimed at enhancing diffusion models, including sampling acceleration and the design of new diffusion processes. In addition, we delve into strategies for implementing diffusion models in manifold and discrete spaces, maximum likelihood training for diffusion models, and methods for creating bridges between two arbitrary distributions. The innovations we discuss represent the efforts for improving the functionality and efficiency of diffusion models in recent years. To examine the efficacy of existing models, a benchmark of FID score, IS, and NLL is presented in a specific NFE. Furthermore, diffusion models are found to be useful in various domains such as computer vision, audio, sequence modeling, and AI for science. The paper concludes with a summary of this field, along with existing limitations and future directions. Summation of existing wellclassified methods is in our Github httpsgithub.comchq1155ASurveyonGenerativeDiffusionModel
Generative Visual Prompt Unifying Distributional Control of PreTrained Generative Models ; Generative models e.g., GANs, diffusion models learn the underlying data distribution in an unsupervised manner. However, many applications of interest require sampling from a particular region of the output space or sampling evenly over a range of characteristics. For efficient sampling in these scenarios, we propose Generative Visual Prompt PromptGen, a framework for distributional control over pretrained generative models by incorporating knowledge of other offtheshelf models. PromptGen defines control as energybased models EBMs and samples images in a feedforward manner by approximating the EBM with invertible neural networks, avoiding optimization at inference. Our experiments demonstrate how PromptGen can efficiently sample from several unconditional generative models e.g., StyleGAN2, StyleNeRF, diffusion autoencoder, NVAE in a controlled orand debiased manner using various offtheshelf models 1 with the CLIP model as control, PromptGen can sample images guided by text, 2 with image classifiers as control, PromptGen can debias generative models across a set of attributes or attribute combinations, and 3 with inverse graphics models as control, PromptGen can sample images of the same identity in different poses. 4 Finally, PromptGen reveals that the CLIP model shows a reporting bias when used as control, and PromptGen can further debias this controlled distribution in an iterative manner. The code is available at httpsgithub.comChenWu98GenerativeVisualPrompt.
Dual Student Networks for DataFree Model Stealing ; Existing datafree model stealing methods use a generator to produce samples in order to train a student model to match the target model outputs. To this end, the two main challenges are estimating gradients of the target model without access to its parameters, and generating a diverse set of training samples that thoroughly explores the input space. We propose a Dual Student method where two students are symmetrically trained in order to provide the generator a criterion to generate samples that the two students disagree on. On one hand, disagreement on a sample implies at least one student has classified the sample incorrectly when compared to the target model. This incentive towards disagreement implicitly encourages the generator to explore more diverse regions of the input space. On the other hand, our method utilizes gradients of student models to indirectly estimate gradients of the target model. We show that this novel training objective for the generator network is equivalent to optimizing a lower bound on the generator's loss if we had access to the target model gradients. We show that our new optimization framework provides more accurate gradient estimation of the target model and better accuracies on benchmark classification datasets. Additionally, our approach balances improved query efficiency with training computation cost. Finally, we demonstrate that our method serves as a better proxy model for transferbased adversarial attacks than existing datafree model stealing methods.
Bottlenecks CLUB Unifying InformationTheoretic Tradeoffs Among Complexity, Leakage, and Utility ; Bottleneck problems are an important class of optimization problems that have recently gained increasing attention in the domain of machine learning and information theory. They are widely used in generative models, fair machine learning algorithms, design of privacyassuring mechanisms, and appear as informationtheoretic performance bounds in various multiuser communication problems. In this work, we propose a general family of optimization problems, termed as complexityleakageutility bottleneck CLUB model, which i provides a unified theoretical framework that generalizes most of the stateoftheart literature for the informationtheoretic privacy models, ii establishes a new interpretation of the popular generative and discriminative models, iii constructs new insights to the generative compression models, and iv can be used in the fair generative models. We first formulate the CLUB model as a complexityconstrained privacyutility optimization problem. We then connect it with the closely related bottleneck problems, namely information bottleneck IB, privacy funnel PF, deterministic IB DIB, conditional entropy bottleneck CEB, and conditional PF CPF. We show that the CLUB model generalizes all these problems as well as most other informationtheoretic privacy models. Then, we construct the deep variational CLUB DVCLUB models by employing neural networks to parameterize variational approximations of the associated information quantities. Building upon these information quantities, we present unified objectives of the supervised and unsupervised DVCLUB models. Leveraging the DVCLUB model in an unsupervised setup, we then connect it with stateoftheart generative models, such as variational autoencoders VAEs, generative adversarial networks GANs, as well as the Wasserstein GAN WGAN, Wasserstein autoencoder WAE, and adversarial autoencoder AAE models through the optimal transport OT problem. We then show that the DVCLUB model can also be used in fair representation learning problems, where the goal is to mitigate the undesired bias during the training phase of a machine learning model. We conduct extensive quantitative experiments on coloredMNIST and CelebA datasets, with a public implementation available, to evaluate and analyze the CLUB model.
Toward a Generalization Metric for Deep Generative Models ; Measuring the generalization capacity of Deep Generative Models DGMs is difficult because of the curse of dimensionality. Evaluation metrics for DGMs such as Inception Score, Fr'echet Inception Distance, PrecisionRecall, and Neural Net Divergence try to estimate the distance between the generated distribution and the target distribution using a polynomial number of samples. These metrics are the target of researchers when designing new models. Despite the claims, it is still unclear how well can they measure the generalization capacity of a generative model. In this paper, we investigate the capacity of these metrics in measuring the generalization capacity. We introduce a framework for comparing the robustness of evaluation metrics. We show that better scores in these metrics do not imply better generalization. They can be fooled easily by a generator that memorizes a small subset of the training set. We propose a fix to the NND metric to make it more robust to noise in the generated data. Toward building a robust metric for generalization, we propose to apply the Minimum Description Length principle to the problem of evaluating DGMs. We develop an efficient method for estimating the complexity of Generative Latent Variable Models GLVMs. Experimental results show that our metric can effectively detect training set memorization and distinguish GLVMs of different generalization capacities. Source code is available at httpsgithub.comhtt210GeneralizationMetricGAN.
Neural Text Generation A Practical Guide ; Deep learning methods have recently achieved great empirical success on machine translation, dialogue response generation, summarization, and other text generation tasks. At a high level, the technique has been to train endtoend neural network models consisting of an encoder model to produce a hidden representation of the source text, followed by a decoder model to generate the target. While such models have significantly fewer pieces than earlier systems, significant tuning is still required to achieve good performance. For text generation models in particular, the decoder can behave in undesired ways, such as by generating truncated or repetitive outputs, outputting bland and generic responses, or in some cases producing ungrammatical gibberish. This paper is intended as a practical guide for resolving such undesired behavior in text generation models, with the aim of helping enable realworld applications.
Generative models with kernel distance in data space ; Generative models dealing with modeling ajoint data distribution are generally either autoencoder or GAN based. Both have their pros and cons, generating blurry images or being unstable in training or prone to mode collapse phenomenon, respectively. The objective of this paper is to construct amodel situated between above architectures, one that does not inherit their main weaknesses. The proposed LCW generator Latent CramerWold generator resembles a classical GAN in transforming Gaussian noise into data space. What is of utmost importance, instead of adiscriminator, LCW generator uses kernel distance. No adversarial training is utilized, hence the name generator. It is trained in two phases. First, an autoencoder based architecture, using kernel measures, is built to model a manifold of data. We propose a Latent Trick mapping a Gaussian to latent in order to get the final model. This results in very competitive FID values.
TwoStreamVAN Improving Motion Modeling in Video Generation ; Video generation is an inherently challenging task, as it requires modeling realistic temporal dynamics as well as spatial content. Existing methods entangle the two intrinsically different tasks of motion and content creation in a single generator network, but this approach struggles to simultaneously generate plausible motion and content. To improve motion modeling in video generation tasks, we propose a twostream model that disentangles motion generation from content generation, called a TwoStream Variational Adversarial Network TwoStreamVAN. Given an action label and a noise vector, our model is able to create clear and consistent motion, and thus yields photorealistic videos. The key idea is to progressively generate and fuse multiscale motion with its corresponding spatial content. Our model significantly outperforms existing methods on the standard Weizmann Human Action, MUG Facial Expression, and VoxCeleb datasets, as well as our new dataset of diverse human actions with challenging and complex motion. Our code is available at httpsgithub.comsunxm2357TwoStreamVAN.
Long and Diverse Text Generation with Planningbased Hierarchical Variational Model ; Existing neural methods for datatotext generation are still struggling to produce long and diverse texts they are insufficient to model input data dynamically during generation, to capture intersentence coherence, or to generate diversified expressions. To address these issues, we propose a Planningbased Hierarchical Variational Model PHVM. Our model first plans a sequence of groups each group is a subset of input items to be covered by a sentence and then realizes each sentence conditioned on the planning result and the previously generated context, thereby decomposing long text generation into dependent sentence generation subtasks. To capture expression diversity, we devise a hierarchical latent structure where a global planning latent variable models the diversity of reasonable planning and a sequence of local latent variables controls sentence realization. Experiments show that our model outperforms stateoftheart baselines in long and diverse text generation.
DeepCopy Grounded Response Generation with Hierarchical Pointer Networks ; Recent advances in neural sequencetosequence models have led to promising results for several language generationbased tasks, including dialogue response generation, summarization, and machine translation. However, these models are known to have several problems, especially in the context of chitchat based dialogue systems they tend to generate short and dull responses that are often too generic. Furthermore, these models do not ground conversational responses on knowledge and facts, resulting in turns that are not accurate, informative and engaging for the users. In this paper, we propose and experiment with a series of response generation models that aim to serve in the general scenario where in addition to the dialogue context, relevant unstructured external knowledge in the form of text is also assumed to be available for models to harness. Our proposed approach extends pointergenerator networks See et al., 2017 by allowing the decoder to hierarchically attend and copy from external knowledge in addition to the dialogue context. We empirically show the effectiveness of the proposed model compared to several baselines including Ghazvininejad et al., 2018; Zhang et al., 2018 through both automatic evaluation metrics and human evaluation on CONVAI2 dataset.
Goaldirected Generation of Discrete Structures with Conditional Generative Models ; Despite recent advances, goaldirected generation of structured discrete data remains challenging. For problems such as program synthesis generating source code and materials design generating molecules, finding examples which satisfy desired constraints or exhibit desired properties is difficult. In practice, expensive heuristic search or reinforcement learning algorithms are often employed. In this paper we investigate the use of conditional generative models which directly attack this inverse problem, by modeling the distribution of discrete structures given properties of interest. Unfortunately, maximum likelihood training of such models often fails with the samples from the generative model inadequately respecting the input properties. To address this, we introduce a novel approach to directly optimize a reinforcement learning objective, maximizing an expected reward. We avoid highvariance scorefunction estimators that would otherwise be required by sampling from an approximation to the normalized rewards, allowing simple Monte Carlo estimation of model gradients. We test our methodology on two tasks generating molecules with userdefined properties and identifying short python expressions which evaluate to a given target value. In both cases, we find improvements over maximum likelihood estimation and other baselines.
Image to Image Translation Generating maps from satellite images ; Generation of maps from satellite images is conventionally done by a range of tools. Maps became an important part of life whose conversion from satellite images may be a bit expensive but Generative models can pander to this challenge. These models aims at finding the patterns between the input and output image. Image to image translation is employed to convert satellite image to corresponding map. Different techniques for image to image translations like Generative adversarial network, Conditional adversarial networks and CoVariational Auto encoders are used to generate the corresponding humanreadable maps for that region, which takes a satellite image at a given zoom level as its input. We are training our model on Conditional Generative Adversarial Network which comprises of Generator model which which generates fake images while the discriminator tries to classify the image as real or fake and both these models are trained synchronously in adversarial manner where both try to fool each other and result in enhancing model performance.
Graphine A Dataset for Graphaware Terminology Definition Generation ; Precisely defining the terminology is the first step in scientific communication. Developing neural text generation models for definition generation can circumvent the laborintensity curation, further accelerating scientific discovery. Unfortunately, the lack of largescale terminology definition dataset hinders the process toward definition generation. In this paper, we present a largescale terminology definition dataset Graphine covering 2,010,648 terminology definition pairs, spanning 227 biomedical subdisciplines. Terminologies in each subdiscipline further form a directed acyclic graph, opening up new avenues for developing graphaware text generation models. We then proposed a novel graphaware definition generation model Graphex that integrates transformer with graph neural network. Our model outperforms existing text generation models by exploiting the graph structure of terminologies. We further demonstrated how Graphine can be used to evaluate pretrained language models, compare graph representation learning methods and predict sentence granularity. We envision Graphine to be a unique resource for definition generation and many other NLP tasks in biomedicine.
In Search of Probeable Generalization Measures ; Understanding the generalization behaviour of deep neural networks is a topic of recent interest that has driven the production of many studies, notably the development and evaluation of generalization explainability measures that quantify model generalization ability. Generalization measures have also proven useful in the development of powerful layerwise model tuning and optimization algorithms, though these algorithms require specific kinds of generalization measures which can probe individual layers. The purpose of this paper is to explore the neglected subtopic of probeable generalization measures; to establish firm ground for further investigations, and to inspire and guide the development of novel model tuning and optimization algorithms. We evaluate and compare measures, demonstrating effectiveness and robustness across model variations, dataset complexities, training hyperparameters, and training stages. We also introduce a new dataset of trained models and performance metrics, GenProb, for testing generalization measures, model tuning algorithms and optimization algorithms.
A Contextual Latent Space Model Subsequence Modulation in Melodic Sequence ; Some generative models for sequences such as music and text allow us to edit only subsequences, given surrounding context sequences, which plays an important part in steering generation interactively. However, editing subsequences mainly involves randomly resampling subsequences from a possible generation space. We propose a contextual latent space model CLSM in order for users to be able to explore subsequence generation with a sense of direction in the generation space, e.g., interpolation, as well as exploring variations semantically similar possible subsequences. A contextinformed prior and decoder constitute the generative model of CLSM, and a context positioninformed encoder is the inference model. In experiments, we use a monophonic symbolic music dataset, demonstrating that our contextual latent space is smoother in interpolation than baselines, and the quality of generated samples is superior to baseline models. The generation examples are available online.
Context Matters in Semantically Controlled Language Generation for Taskoriented Dialogue Systems ; This work combines information about the dialogue history encoded by pretrained model with a meaning representation of the current system utterance to realize contextual language generation in taskoriented dialogues. We utilize the pretrained multicontext ConveRT model for context representation in a model trained from scratch; and leverage the immediate preceding user utterance for context generation in a model adapted from the pretrained GPT2. Both experiments with the MultiWOZ dataset show that contextual information encoded by pretrained model improves the performance of response generation both in automatic metrics and human evaluation. Our presented contextual generator enables higher variety of generated responses that fit better to the ongoing dialogue. Analysing the context size shows that longer context does not automatically lead to better performance, but the immediate preceding user utterance plays an essential role for contextual generation. In addition, we also propose a reranker for the GPTbased generation model. The experiments show that the response selected by the reranker has a significant improvement on automatic metrics.
Physics guided deep learning generative models for crystal materials discovery ; Deep learning based generative models such as deepfake have been able to generate amazing images and videos. However, these models may need significant transformation when applied to generate crystal materials structures in which the building blocks, the physical atoms are very different from the pixels. Naively transferred generative models tend to generate a large portion of physically infeasible crystal structures that are not stable or synthesizable. Herein we show that by exploiting and adding physically oriented data augmentation, loss function terms, and post processing, our deep adversarial network GAN based generative models can now generate crystal structures with higher physical feasibility and expand our previous models which can only create cubic structures.
GoalDirected Story Generation Augmenting Generative Language Models with Reinforcement Learning ; The advent of large pretrained generative language models has provided a common framework for AI story generation via sampling the model to create sequences that continue the story. However, sampling alone is insufficient for story generation. In particular, it is hard to direct a language model to create stories to reach a specific goal event. We present two automated techniques grounded in deep reinforcement learning and reward shaping to control the plot of computergenerated stories. The first utilizes proximal policy optimization to finetune an existing transformerbased language model to generate text continuations but also be goalseeking. The second extracts a knowledge graph from the unfolding story, which is used by a policy network with graph attention to select a candidate continuation generated by a language model. We report on automated metrics pertaining to how often stories achieve a given goal event as well as human participant rankings of coherence and overall story quality compared to baselines and ablations.
Training and Tuning Generative Neural Radiance Fields for AttributeConditional 3DAware Face Generation ; 3Daware GANs based on generative neural radiance fields GNeRF have achieved impressive highquality image generation, while preserving strong 3D consistency. The most notable achievements are made in the face generation domain. However, most of these models focus on improving view consistency but neglect a disentanglement aspect, thus these models cannot provide highquality semanticattribute control over generation. To this end, we introduce a conditional GNeRF model that uses specific attribute labels as input in order to improve the controllabilities and disentangling abilities of 3Daware generative models. We utilize the pretrained 3Daware model as the basis and integrate a dualbranches attributeediting module DAEM, that utilize attribute labels to provide control over generation. Moreover, we propose a TRIOT TRaining as Init, and Optimizing for Tuning method to optimize the latent vector to improve the precision of the attributeediting further. Extensive experiments on the widely used FFHQ show that our model yields highquality editing with better view consistency while preserving the nontarget regions. The code is available at httpsgithub.comzhangqianhuiTTGNeRF.
MultiSource Diffusion Models for Simultaneous Music Generation and Separation ; In this work, we define a diffusionbased generative model capable of both music synthesis and source separation by learning the score of the joint probability density of sources sharing a context. Alongside the classic total inference tasks i.e., generating a mixture, separating the sources, we also introduce and experiment on the partial generation task of source imputation, where we generate a subset of the sources given the others e.g., play a piano track that goes well with the drums. Additionally, we introduce a novel inference method for the separation task based on Dirac likelihood functions. We train our model on Slakh2100, a standard dataset for musical source separation, provide qualitative results in the generation settings, and showcase competitive quantitative results in the source separation setting. Our method is the first example of a single model that can handle both generation and separation tasks, thus representing a step toward general audio models.
Diffusing Gaussian Mixtures for Generating Categorical Data ; Learning a categorical distribution comes with its own set of challenges. A successful approach taken by stateoftheart works is to cast the problem in a continuous domain to take advantage of the impressive performance of the generative models for continuous data. Amongst them are the recently emerging diffusion probabilistic models, which have the observed advantage of generating highquality samples. Recent advances for categorical generative models have focused on log likelihood improvements. In this work, we propose a generative model for categorical data based on diffusion models with a focus on highquality sample generation, and propose sampledbased evaluation methods. The efficacy of our method stems from performing diffusion in the continuous domain while having its parameterization informed by the structure of the categorical nature of the target distribution. Our method of evaluation highlights the capabilities and limitations of different generative models for generating categorical data, and includes experiments on synthetic and realworld protein datasets.
MakeAnAnimation LargeScale Textconditional 3D Human Motion Generation ; Textguided human motion generation has drawn significant interest because of its impactful applications spanning animation and robotics. Recently, application of diffusion models for motion generation has enabled improvements in the quality of generated motions. However, existing approaches are limited by their reliance on relatively smallscale motion capture data, leading to poor performance on more diverse, inthewild prompts. In this paper, we introduce MakeAnAnimation, a textconditioned human motion generation model which learns more diverse poses and prompts from largescale imagetext datasets, enabling significant improvement in performance over prior works. MakeAnAnimation is trained in two stages. First, we train on a curated largescale dataset of text, static pseudopose pairs extracted from imagetext datasets. Second, we finetune on motion capture data, adding additional layers to model the temporal dimension. Unlike prior diffusion models for motion generation, MakeAnAnimation uses a UNet architecture similar to recent texttovideo generation models. Human evaluation of motion realism and alignment with input text shows that our model reaches stateoftheart performance on texttomotion generation.
SelfConsuming Generative Models Go MAD ; Seismic advances in generative AI algorithms for imagery, text, and other data types has led to the temptation to use synthetic data to train nextgeneration models. Repeating this process creates an autophagous selfconsuming loop whose properties are poorly understood. We conduct a thorough analytical and empirical analysis using stateoftheart generative image models of three families of autophagous loops that differ in how fixed or fresh real training data is available through the generations of training and in whether the samples from previous generation models have been biased to trade off data quality versus diversity. Our primary conclusion across all scenarios is that without enough fresh real data in each generation of an autophagous loop, future generative models are doomed to have their quality precision or diversity recall progressively decrease. We term this condition Model Autophagy Disorder MAD, making analogy to mad cow disease.
Deep Generative Models, Synthetic Tabular Data, and Differential Privacy An Overview and Synthesis ; This article provides a comprehensive synthesis of the recent developments in synthetic data generation via deep generative models, focusing on tabular datasets. We specifically outline the importance of synthetic data generation in the context of privacysensitive data. Additionally, we highlight the advantages of using deep generative models over other methods and provide a detailed explanation of the underlying concepts, including unsupervised learning, neural networks, and generative models. The paper covers the challenges and considerations involved in using deep generative models for tabular datasets, such as data normalization, privacy concerns, and model evaluation. This review provides a valuable resource for researchers and practitioners interested in synthetic data generation and its applications.
A meanfield games laboratory for generative modeling ; In this paper, we demonstrate the versatility of meanfield games MFGs as a mathematical framework for explaining, enhancing, and designing generative models. There is a pervasive sense in the generative modeling community that the various flow and diffusionbased generative models have some common foundational structure and interrelationships. We establish connections between MFGs and major classes of flow and diffusionbased generative models including continuoustime normalizing flows, scorebased models, and Wasserstein gradient flows. We derive these three classes of generative models through different choices of particle dynamics and cost functions. Furthermore, we study the mathematical structure and properties of each generative model by studying their associated MFG's optimality condition, which is a set of coupled forwardbackward nonlinear partial differential equations PDEs. The theory of MFGs, therefore, enables the study of generative models through the theory of nonlinear PDEs. Through this perspective, we investigate the wellposedness and structure of normalizing flows, unravel the mathematical structure of scorebased generative modeling, and derive a meanfield game formulation of the Wasserstein gradient flow. From an algorithmic perspective, the optimality conditions of MFGs also allow us to introduce HJB regularizers for enhanced training of a broad class of generative models. In particular, we propose and demonstrate an HamiltonJacobiBellman regularized SGM with improved performance over standard SGMs. We present this framework as an MFG laboratory which serves as a platform for revealing new avenues of experimentation and invention of generative models. This laboratory will give rise to a multitude of wellposed generative modeling formulations and will provide a consistent theoretical framework upon which numerical and algorithmic tools may be developed.
Generalizing the generalized Chaplygin gas ; The generalized Chaplygin gas is characterized by the equation of state p Arhoalpha, with alpha 1 and w 1. We generalize this model to allow for the cases where alpha 1 or w 1. This generalization leads to three new versions of the generalized Chaplygin gas an early phantom model in which w 1 at early times and asymptotically approaches w 1 at late times, a late phantom model with w approx 1 at early times and w infty at late times, and a transient model with w approx 1 at early times and w 0 at late times. We consider these three cases as models for dark energy alone and examine constraints from type Ia supernovae and from the subhorizon growth of density perturbations. The transient Chaplygin gas model provides a possible mechanism to allow for a currently accelerating universe without a future horizon, while some of the early phantom models produce w 1 without either past or future singularities.
Generation Mass Hierarchy in Superstring Derived Models ; I discuss the problem of generation mass hierarchy in the context of realistic superstring models which are constructed in the free fermionic formulation. These models correspond to models which are compactified on Z2times Z2 orbifold. I suggest that the hierarchy among the generations results from horizontal symmetries, which arise from the compactification. In particular, I show that in a class of free fermionic standardlike models, the suppression of the mass terms for the lightest generation is a general, and unambiguous, characteristic of these models. I show that the mixing between the generations is suppressed due to the horizontal symmetries. I conclude that these models may potentially explain the generation mass hierarchy.
Emergent general relativity in the tensor models possessing Gaussian classical solutions ; This paper gives a summary of the author's works concerning the emergent general relativity in a particular class of tensor models, which possess Gaussian classical solutions. In general, a classical solution in a tensor model may be physically regarded as a background space, and small fluctuations about the solution as emergent fields on the space. The numerical analyses of the tensor models possessing Gaussian classical background solutions have shown that the lowlying longwavelength fluctuations around the backgrounds are in onetoone correspondence with the geometric fluctuations on flat spaces in the general relativity. It has also been shown that part of the orthogonal symmetry of the tensor model spontaneously broken by the backgrounds can be identified with the local translation symmetry of the general relativity. Thus the tensor model provides an interesting model of simultaneous emergence of space, the general relativity, and its local gauge symmetry of translation.
Generative Model Selection Using a Scalable and SizeIndependent Complex Network Classifier ; Real networks exhibit nontrivial topological features such as heavytailed degree distribution, high clustering, and smallworldness. Researchers have developed several generative models for synthesizing artificial networks that are structurally similar to real networks. An important research problem is to identify the generative model that best fits to a target network. In this paper, we investigate this problem and our goal is to select the model that is able to generate graphs similar to a given network instance. By the means of generating synthetic networks with seven outstanding generative models, we have utilized machine learning methods to develop a decision tree for model selection. Our proposed method, which is named Generative Model Selection for Complex Networks GMSCN, outperforms existing methods with respect to accuracy, scalability and sizeindependence.
Improving image generative models with human interactions ; GANs provide a framework for training generative models which mimic a data distribution. However, in many cases we wish to train these generative models to optimize some auxiliary objective function within the data it generates, such as making more aesthetically pleasing images. In some cases, these objective functions are difficult to evaluate, e.g. they may require human interaction. Here, we develop a system for efficiently improving a GAN to target an objective involving human interaction, specifically generating images that increase rates of positive user interactions. To improve the generative model, we build a model of human behavior in the targeted domain from a relatively small set of interactions, and then use this behavioral model as an auxiliary loss function to improve the generative model. We show that this system is successful at improving positive interaction rates, at least on simulated data, and characterize some of the factors that affect its performance.
Kernel Mean Matching for Content Addressability of GANs ; We propose a novel procedure which adds contentaddressability to any given unconditional implicit model e.g., a generative adversarial network GAN. The procedure allows users to control the generative process by specifying a set arbitrary size of desired examples based on which similar samples are generated from the model. The proposed approach, based on kernel mean matching, is applicable to any generative models which transform latent vectors to samples, and does not require retraining of the model. Experiments on various highdimensional image generation problems CelebAHQ, LSUN bedroom, bridge, tower show that our approach is able to generate images which are consistent with the input set, while retaining the image quality of the original model. To our knowledge, this is the first work that attempts to construct, at test time, a contentaddressable generative model from a trained marginal model.
A general model for planebased clustering with loss function ; In this paper, we propose a general model for planebased clustering. The general model contains many existing planebased clustering methods, e.g., kplane clustering kPC, proximal plane clustering PPC, twin support vector clustering TWSVC and its extensions. Under this general model, one may obtain an appropriate clustering method for specific purpose. The general model is a procedure corresponding to an optimization problem, where the optimization problem minimizes the total loss of the samples. Thereinto, the loss of a sample derives from both withincluster and betweencluster. In theory, the termination conditions are discussed, and we prove that the general model terminates in a finite number of steps at a local or weak local optimal point. Furthermore, based on this general model, we propose a planebased clustering method by introducing a new loss function to capture the data distribution precisely. Experimental results on artificial and public available datasets verify the effectiveness of the proposed method.
Outfit Generation and Style Extraction via Bidirectional LSTM and Autoencoder ; When creating an outfit, style is a criterion in selecting each fashion item. This means that style can be regarded as a feature of the overall outfit. However, in various previous studies on outfit generation, there have been few methods focusing on global information obtained from an outfit. To address this deficiency, we have incorporated an unsupervised style extraction module into a model to learn outfits. Using the style information of an outfit as a whole, the proposed model succeeded in generating outfits more flexibly without requiring additional information. Moreover, the style information extracted by the proposed model is easy to interpret. The proposed model was evaluated on two humangenerated outfit datasets. In a fashion item prediction task missing prediction task, the proposed model outperformed a baseline method. In a style extraction task, the proposed model extracted some easily distinguishable styles. In an outfit generation task, the proposed model generated an outfit while controlling its styles. This capability allows us to generate fashionable outfits according to various preferences.
What leads to generalization of object proposals ; Object proposal generation is often the first step in many detection models. It is lucrative to train a good proposal model, that generalizes to unseen classes. This could help scaling detection models to larger number of classes with fewer annotations. Motivated by this, we study how a detection model trained on a small set of source classes can provide proposals that generalize to unseen classes. We systematically study the properties of the dataset visual diversity and label space granularity required for good generalization. We show the tradeoff between using finegrained labels and coarse labels. We introduce the idea of prototypical classes a set of sufficient and necessary classes required to train a detection model to obtain generalized proposals in a more dataefficient way. On the Open Images V4 dataset, we show that only 25 of the classes can be selected to form such a prototypical set. The resulting proposals from a model trained with these classes is only 4.3 worse than using all the classes, in terms of average recall AR. We also demonstrate that Faster RCNN model leads to better generalization of proposals compared to a singlestage network like RetinaNet.
Learning Neural Generative Dynamics for Molecular Conformation Generation ; We study how to generate molecule conformations i.e., 3D structures from a molecular graph. Traditional methods, such as molecular dynamics, sample conformations via computationally expensive simulations. Recently, machine learning methods have shown great potential by training on a large collection of conformation data. Challenges arise from the limited model capacity for capturing complex distributions of conformations and the difficulty in modeling longrange dependencies between atoms. Inspired by the recent progress in deep generative models, in this paper, we propose a novel probabilistic framework to generate valid and diverse conformations given a molecular graph. We propose a method combining the advantages of both flowbased and energybased models, enjoying 1 a high model capacity to estimate the multimodal conformation distribution; 2 explicitly capturing the complex longrange dependencies between atoms in the observation space. Extensive experiments demonstrate the superior performance of the proposed method on several benchmarks, including conformation generation and distance modeling tasks, with a significant improvement over existing generative models for molecular conformation sampling.
A Discrete CVAE for Response Generation on ShortText Conversation ; Neural conversation models such as encoderdecoder models are easy to generate bland and generic responses. Some researchers propose to use the conditional variational autoencoderCVAE which maximizes the lower bound on the conditional loglikelihood on a continuous latent variable. With different sampled latent variables, the model is expected to generate diverse responses. Although the CVAEbased models have shown tremendous potential, their improvement of generating highquality responses is still unsatisfactory. In this paper, we introduce a discrete latent variable with an explicit semantic meaning to improve the CVAE on shorttext conversation. A major advantage of our model is that we can exploit the semantic distance between the latent variables to maintain good diversity between the sampled latent variables. Accordingly, we propose a twostage sampling approach to enable efficient diverse variable selection from a large latent space assumed in the shorttext conversation task. Experimental results indicate that our model outperforms various kinds of generation models under both automatic and human evaluations and generates more diverse and informative responses.
TexttoImage Generation with Attention Based Recurrent Neural Networks ; Conditional image modeling based on textual descriptions is a relatively new domain in unsupervised learning. Previous approaches use a latent variable model and generative adversarial networks. While the formers are approximated by using variational autoencoders and rely on the intractable inference that can hamper their performance, the latter is unstable to train due to Nash equilibrium based objective function. We develop a tractable and stable captionbased image generation model. The model uses an attentionbased encoder to learn wordtopixel dependencies. A conditional autoregressive based decoder is used for learning pixeltopixel dependencies and generating images. Experimentations are performed on Microsoft COCO, and MNISTwithcaptions datasets and performance is evaluated by using the Structural Similarity Index. Results show that the proposed model performs better than contemporary approaches and generate better quality images. Keywords Generative image modeling, autoregressive image modeling, captionbased image generation, neural attention, recurrent neural networks.
Automated Diagram Generation to Build Understanding and Usability ; Causal loop and stock and flow diagrams are broadly used in System Dynamics because they help organize relationships and convey meaning. Using the analytical work of Schoenberg 2019 to select what to include in a compressed model, this paper demonstrates how that information can be clearly presented in an automatically generated causal loop diagram. The diagrams are generated using tools developed by people working in graph theory and the generated diagrams are clear and aesthetically pleasing. This approach can also be built upon to generate stock and flow diagrams. Automated stock and flow diagram generation opens the door to representing models developed using only equations, regardless or origin, in a clear and easy to understand way. Because models can be large, the application of grouping techniques, again developed for graph theory, can help structure the resulting diagrams in the most usable form. This paper describes the algorithms developed for automated diagram generation and shows a number of examples of their uses in large models. The application of these techniques to existing, but inaccessible, equationbased models can help broaden the knowledge base for System Dynamics modeling. The techniques can also be used to improve layout in all, or part, of existing models with diagrammatic informtion.
Conditional Generative Models for Counterfactual Explanations ; Counterfactual instances offer humaninterpretable insight into the local behaviour of machine learning models. We propose a general framework to generate sparse, indistribution counterfactual model explanations which match a desired target prediction with a conditional generative model, allowing batches of counterfactual instances to be generated with a single forward pass. The method is flexible with respect to the type of generative model used as well as the task of the underlying predictive model. This allows straightforward application of the framework to different modalities such as images, time series or tabular data as well as generative model paradigms such as GANs or autoencoders and predictive tasks like classification or regression. We illustrate the effectiveness of our method on image CelebA, time series ECG and mixedtype tabular Adult Census data.
Topical Language Generation using Transformers ; Largescale transformerbased language models LMs demonstrate impressive capabilities in open text generation. However, controlling the generated text's properties such as the topic, style, and sentiment is challenging and often requires significant changes to the model architecture or retraining and finetuning the model on new supervised data. This paper presents a novel approach for Topical Language Generation TLG by combining a pretrained LM with topic modeling information. We cast the problem using Bayesian probability formulation with topic probabilities as a prior, LM probabilities as the likelihood, and topical language generation probability as the posterior. In learning the model, we derive the topic probability distribution from the userprovided document's natural structure. Furthermore, we extend our model by introducing new parameters and functions to influence the quantity of the topical features presented in the generated text. This feature would allow us to easily control the topical properties of the generated text. Our experimental results demonstrate that our model outperforms the stateoftheart results on coherency, diversity, and fluency while being faster in decoding.