Wenlin Wang

Qian Yang, Zhouyuan Huo, Wenlin Wang et al.

Language models are essential for natural language processing (NLP) tasks, such as machine translation and text summarization. Remarkable performance has been demonstrated recently across many NLP domains via a Transformer-based language model with over a billion parameters, verifying the benefits of model size. Model parallelism is required if a model is too large to fit in a single computing device. Current methods for model parallelism either suffer from backward locking in backpropagation or are not applicable to language models. We propose the first model-parallel algorithm that speeds the training of Transformer-based language models. We also prove that our proposed algorithm is guaranteed to converge to critical points for non-convex problems. Extensive experiments on Transformer and Transformer-XL language models demonstrate that the proposed algorithm obtains a much faster speedup beyond data parallelism, with comparable or better accuracy. Code to reproduce experiments is to be found at \url{https://github.com/LaraQianYang/Ouroboros}.

Dinghan Shen, Guoyin Wang, Wenlin Wang et al.

Many deep learning architectures have been proposed to model the compositionality in text sequences, requiring a substantial number of parameters and expensive computations. However, there has not been a rigorous evaluation regarding the added value of sophisticated compositional functions. In this paper, we conduct a point-by-point comparative study between Simple Word-Embedding-based Models (SWEMs), consisting of parameter-free pooling operations, relative to word-embedding-based RNN/CNN models. Surprisingly, SWEMs exhibit comparable or even superior performance in the majority of cases considered. Based upon this understanding, we propose two additional pooling strategies over learned word embeddings: (i) a max-pooling operation for improved interpretability; and (ii) a hierarchical pooling operation, which preserves spatial (n-gram) information within text sequences. We present experiments on 17 datasets encompassing three tasks: (i) (long) document classification; (ii) text sequence matching; and (iii) short text tasks, including classification and tagging. The source code and datasets can be obtained from https:// github.com/dinghanshen/SWEM.

Hao Liu, Qian Gao, Jiang Li et al.

In modern internet industries, deep learning based recommender systems have became an indispensable building block for a wide spectrum of applications, such as search engine, news feed, and short video clips. However, it remains challenging to carry the well-trained deep models for online real-time inference serving, with respect to the time-varying web-scale traffics from billions of users, in a cost-effective manner. In this work, we present JIZHI - a Model-as-a-Service system - that per second handles hundreds of millions of online inference requests to huge deep models with more than trillions of sparse parameters, for over twenty real-time recommendation services at Baidu, Inc. In JIZHI, the inference workflow of every recommendation request is transformed to a Staged Event-Driven Pipeline (SEDP), where each node in the pipeline refers to a staged computation or I/O intensive task processor. With traffics of real-time inference requests arrived, each modularized processor can be run in a fully asynchronized way and managed separately. Besides, JIZHI introduces heterogeneous and hierarchical storage to further accelerate the online inference process by reducing unnecessary computations and potential data access latency induced by ultra-sparse model parameters. Moreover, an intelligent resource manager has been deployed to maximize the throughput of JIZHI over the shared infrastructure by searching the optimal resource allocation plan from historical logs and fine-tuning the load shedding policies over intermediate system feedback. Extensive experiments have been done to demonstrate the advantages of JIZHI from the perspectives of end-to-end service latency, system-wide throughput, and resource consumption. JIZHI has helped Baidu saved more than ten million US dollars in hardware and utility costs while handling 200% more traffics without sacrificing inference efficiency.

Guoyin Wang, Chunyuan Li, Jianqiao Li et al.

Neural language models are often trained with maximum likelihood estimation (MLE), where the next word is generated conditioned on the ground-truth word tokens. During testing, however, the model is instead conditioned on previously generated tokens, resulting in what is termed exposure bias. To reduce this gap between training and testing, we propose using optimal transport (OT) to match the sequences generated in these two modes. An extension is further proposed to improve the OT learning, based on the structural and contextual information of the text sequences. The effectiveness of the proposed method is validated on machine translation, text summarization, and text generation tasks.

Ruiyi Zhang, Changyou Chen, Zhe Gan et al.

Auto-regressive text generation models usually focus on local fluency, and may cause inconsistent semantic meaning in long text generation. Further, automatically generating words with similar semantics is challenging, and hand-crafted linguistic rules are difficult to apply. We consider a text planning scheme and present a model-based imitation-learning approach to alleviate the aforementioned issues. Specifically, we propose a novel guider network to focus on the generative process over a longer horizon, which can assist next-word prediction and provide intermediate rewards for generator optimization. Extensive experiments demonstrate that the proposed method leads to improved performance.

Ruiyi Zhang, Changyou Chen, Zhe Gan et al.

Reinforcement learning (RL) has been widely studied for improving sequence-generation models. However, the conventional rewards used for RL training typically cannot capture sufficient semantic information and therefore render model bias. Further, the sparse and delayed rewards make RL exploration inefficient. To alleviate these issues, we propose the concept of nested-Wasserstein distance for distributional semantic matching. To further exploit it, a novel nested-Wasserstein self-imitation learning framework is developed, encouraging the model to exploit historical high-rewarded sequences for enhanced exploration and better semantic matching. Our solution can be understood as approximately executing proximal policy optimization with Wasserstein trust-regions. Experiments on a variety of unconditional and conditional sequence-generation tasks demonstrate the proposed approach consistently leads to improved performance.

Wenlin Wang, Hongteng Xu, Zhe Gan et al.

We propose a novel graph-driven generative model, that unifies multiple heterogeneous learning tasks into the same framework. The proposed model is based on the fact that heterogeneous learning tasks, which correspond to different generative processes, often rely on data with a shared graph structure. Accordingly, our model combines a graph convolutional network (GCN) with multiple variational autoencoders, thus embedding the nodes of the graph i.e., samples for the tasks) in a uniform manner while specializing their organization and usage to different tasks. With a focus on healthcare applications (tasks), including clinical topic modeling, procedure recommendation and admission-type prediction, we demonstrate that our method successfully leverages information across different tasks, boosting performance in all tasks and outperforming existing state-of-the-art approaches.

Wenlin Wang, Hongteng Xu, Ruiyi Zhang et al.

Traditional collaborative filtering (CF) based recommender systems tend to perform poorly when the user-item interactions/ratings are highly scarce. To address this, we propose a learning framework that improves collaborative filtering with a synthetic feedback loop (CF-SFL) to simulate the user feedback. The proposed framework consists of a "recommender" and a "virtual user". The "recommender" is formulated as a CF model, recommending items according to observed user preference. The "virtual user" estimates rewards from the recommended items and generates a \emph{feedback} in addition to the observed user preference. The "recommender" connected with the "virtual user" constructs a closed loop, that recommends users with items and imitates the \emph{unobserved} feedback of the users to the recommended items. The synthetic feedback is used to augment the observed user preference and improve recommendation results. Theoretically, such model design can be interpreted as inverse reinforcement learning, which can be learned effectively via rollout (simulation). Experimental results show that the proposed framework is able to enrich the learning of user preference and boost the performance of existing collaborative filtering methods on multiple datasets.

Wenlin Wang, Hongteng Xu, Guoyin Wang et al.

We propose an optimal transport (OT) framework for generalized zero-shot learning (GZSL), seeking to distinguish samples for both seen and unseen classes, with the assist of auxiliary attributes. The discrepancy between features and attributes is minimized by solving an optimal transport problem. {Specifically, we build a conditional generative model to generate features from seen-class attributes, and establish an optimal transport between the distribution of the generated features and that of the real features.} The generative model and the optimal transport are optimized iteratively with an attribute-based regularizer, that further enhances the discriminative power of the generated features. A classifier is learned based on the features generated for both the seen and unseen classes. In addition to generalized zero-shot learning, our framework is also applicable to standard and transductive ZSL problems. Experiments show that our optimal transport-based method outperforms state-of-the-art methods on several benchmark datasets.

Wenlin Wang, Chenyang Tao, Zhe Gan et al.

The performance of many network learning applications crucially hinges on the success of network embedding algorithms, which aim to encode rich network information into low-dimensional vertex-based vector representations. This paper considers a novel variational formulation of network embeddings, with special focus on textual networks. Different from most existing methods that optimize a discriminative objective, we introduce Variational Homophilic Embedding (VHE), a fully generative model that learns network embeddings by modeling the semantic (textual) information with a variational autoencoder, while accounting for the structural (topology) information through a novel homophilic prior design. Homophilic vertex embeddings encourage similar embedding vectors for related (connected) vertices. The proposed VHE promises better generalization for downstream tasks, robustness to incomplete observations, and the ability to generalize to unseen vertices. Extensive experiments on real-world networks, for multiple tasks, demonstrate that the proposed method consistently achieves superior performance relative to competing state-of-the-art approaches.

Liqun Chen, Guoyin Wang, Chenyang Tao et al.

Constituting highly informative network embeddings is an important tool for network analysis. It encodes network topology, along with other useful side information, into low-dimensional node-based feature representations that can be exploited by statistical modeling. This work focuses on learning context-aware network embeddings augmented with text data. We reformulate the network-embedding problem, and present two novel strategies to improve over traditional attention mechanisms: ($i$) a content-aware sparse attention module based on optimal transport, and ($ii$) a high-level attention parsing module. Our approach yields naturally sparse and self-normalized relational inference. It can capture long-term interactions between sequences, thus addressing the challenges faced by existing textual network embedding schemes. Extensive experiments are conducted to demonstrate our model can consistently outperform alternative state-of-the-art methods.

Bai Li, Changyou Chen, Wenlin Wang et al.

Adversarial examples are carefully perturbed in-puts for fooling machine learning models. A well-acknowledged defense method against such examples is adversarial training, where adversarial examples are injected into training data to increase robustness. In this paper, we propose a new attack to unveil an undesired property of the state-of-the-art adversarial training, that is it fails to obtain robustness against perturbations in $\ell_2$ and $\ell_\infty$ norms simultaneously. We discuss a possible solution to this issue and its limitations as well.

Wenlin Wang, Zhe Gan, Hongteng Xu et al.

We propose a topic-guided variational autoencoder (TGVAE) model for text generation. Distinct from existing variational autoencoder (VAE) based approaches, which assume a simple Gaussian prior for the latent code, our model specifies the prior as a Gaussian mixture model (GMM) parametrized by a neural topic module. Each mixture component corresponds to a latent topic, which provides guidance to generate sentences under the topic. The neural topic module and the VAE-based neural sequence module in our model are learned jointly. In particular, a sequence of invertible Householder transformations is applied to endow the approximate posterior of the latent code with high flexibility during model inference. Experimental results show that our TGVAE outperforms alternative approaches on both unconditional and conditional text generation, which can generate semantically-meaningful sentences with various topics.

Ruiyi Zhang, Changyou Chen, Zhe Gan et al.

Sequence generation with reinforcement learning (RL) has received significant attention recently. However, a challenge with such methods is the sparse-reward problem in the RL training process, in which a scalar guiding signal is often only available after an entire sequence has been generated. This type of sparse reward tends to ignore the global structural information of a sequence, causing generation of sequences that are semantically inconsistent. In this paper, we present a model-based RL approach to overcome this issue. Specifically, we propose a novel guider network to model the sequence-generation environment, which can assist next-word prediction and provide intermediate rewards for generator optimization. Extensive experiments show that the proposed method leads to improved performance for both unconditional and conditional sequence-generation tasks.

Hongteng Xu, Wenlin Wang, Wei Liu et al.

We propose a novel Wasserstein method with a distillation mechanism, yielding joint learning of word embeddings and topics. The proposed method is based on the fact that the Euclidean distance between word embeddings may be employed as the underlying distance in the Wasserstein topic model. The word distributions of topics, their optimal transports to the word distributions of documents, and the embeddings of words are learned in a unified framework. When learning the topic model, we leverage a distilled underlying distance matrix to update the topic distributions and smoothly calculate the corresponding optimal transports. Such a strategy provides the updating of word embeddings with robust guidance, improving the algorithmic convergence. As an application, we focus on patient admission records, in which the proposed method embeds the codes of diseases and procedures and learns the topics of admissions, obtaining superior performance on clinically-meaningful disease network construction, mortality prediction as a function of admission codes, and procedure recommendation.

Bai Li, Changyou Chen, Wenlin Wang et al.

The existence of adversarial data examples has drawn significant attention in the deep-learning community; such data are seemingly minimally perturbed relative to the original data, but lead to very different outputs from a deep-learning algorithm. Although a significant body of work on developing defensive models has been considered, most such models are heuristic and are often vulnerable to adaptive attacks. Defensive methods that provide theoretical robustness guarantees have been studied intensively, yet most fail to obtain non-trivial robustness when a large-scale model and data are present. To address these limitations, we introduce a framework that is scalable and provides certified bounds on the norm of the input manipulation for constructing adversarial examples. We establish a connection between robustness against adversarial perturbation and additive random noise, and propose a training strategy that can significantly improve the certified bounds. Our evaluation on MNIST, CIFAR-10 and ImageNet suggests that the proposed method is scalable to complicated models and large data sets, while providing competitive robustness to state-of-the-art provable defense methods.

Changyou Chen, Ruiyi Zhang, Wenlin Wang et al.

There has been recent interest in developing scalable Bayesian sampling methods such as stochastic gradient MCMC (SG-MCMC) and Stein variational gradient descent (SVGD) for big-data analysis. A standard SG-MCMC algorithm simulates samples from a discrete-time Markov chain to approximate a target distribution, thus samples could be highly correlated, an undesired property for SG-MCMC. In contrary, SVGD directly optimizes a set of particles to approximate a target distribution, and thus is able to obtain good approximations with relatively much fewer samples. In this paper, we propose a principle particle-optimization framework based on Wasserstein gradient flows to unify SG-MCMC and SVGD, and to allow new algorithms to be developed. Our framework interprets SG-MCMC as particle optimization on the space of probability measures, revealing a strong connection between SG-MCMC and SVGD. The key component of our framework is several particle-approximate techniques to efficiently solve the original partial differential equations on the space of probability measures. Extensive experiments on both synthetic data and deep neural networks demonstrate the effectiveness and efficiency of our framework for scalable Bayesian sampling.

Dinghan Shen, Qinliang Su, Paidamoyo Chapfuwa et al.

Semantic hashing has become a powerful paradigm for fast similarity search in many information retrieval systems. While fairly successful, previous techniques generally require two-stage training, and the binary constraints are handled ad-hoc. In this paper, we present an end-to-end Neural Architecture for Semantic Hashing (NASH), where the binary hashing codes are treated as Bernoulli latent variables. A neural variational inference framework is proposed for training, where gradients are directly back-propagated through the discrete latent variable to optimize the hash function. We also draw connections between proposed method and rate-distortion theory, which provides a theoretical foundation for the effectiveness of the proposed framework. Experimental results on three public datasets demonstrate that our method significantly outperforms several state-of-the-art models on both unsupervised and supervised scenarios.

Guoyin Wang, Chunyuan Li, Wenlin Wang et al.

Word embeddings are effective intermediate representations for capturing semantic regularities between words, when learning the representations of text sequences. We propose to view text classification as a label-word joint embedding problem: each label is embedded in the same space with the word vectors. We introduce an attention framework that measures the compatibility of embeddings between text sequences and labels. The attention is learned on a training set of labeled samples to ensure that, given a text sequence, the relevant words are weighted higher than the irrelevant ones. Our method maintains the interpretability of word embeddings, and enjoys a built-in ability to leverage alternative sources of information, in addition to input text sequences. Extensive results on the several large text datasets show that the proposed framework outperforms the state-of-the-art methods by a large margin, in terms of both accuracy and speed.

Wenqi Wang, Yifan Sun, Brian Eriksson et al.

Deep neural networks have demonstrated state-of-the-art performance in a variety of real-world applications. In order to obtain performance gains, these networks have grown larger and deeper, containing millions or even billions of parameters and over a thousand layers. The trade-off is that these large architectures require an enormous amount of memory, storage, and computation, thus limiting their usability. Inspired by the recent tensor ring factorization, we introduce Tensor Ring Networks (TR-Nets), which significantly compress both the fully connected layers and the convolutional layers of deep neural networks. Our results show that our TR-Nets approach {is able to compress LeNet-5 by $11\times$ without losing accuracy}, and can compress the state-of-the-art Wide ResNet by $243\times$ with only 2.3\% degradation in {Cifar10 image classification}. Overall, this compression scheme shows promise in scientific computing and deep learning, especially for emerging resource-constrained devices such as smartphones, wearables, and IoT devices.

Wenlin Wang, Zhe Gan, Wenqi Wang et al.

We propose a Topic Compositional Neural Language Model (TCNLM), a novel method designed to simultaneously capture both the global semantic meaning and the local word ordering structure in a document. The TCNLM learns the global semantic coherence of a document via a neural topic model, and the probability of each learned latent topic is further used to build a Mixture-of-Experts (MoE) language model, where each expert (corresponding to one topic) is a recurrent neural network (RNN) that accounts for learning the local structure of a word sequence. In order to train the MoE model efficiently, a matrix factorization method is applied, by extending each weight matrix of the RNN to be an ensemble of topic-dependent weight matrices. The degree to which each member of the ensemble is used is tied to the document-dependent probability of the corresponding topics. Experimental results on several corpora show that the proposed approach outperforms both a pure RNN-based model and other topic-guided language models. Further, our model yields sensible topics, and also has the capacity to generate meaningful sentences conditioned on given topics.

Kai Fan, Qi Wei, Wenlin Wang et al.

We propose a new method that uses deep learning techniques to solve the inverse problems. The inverse problem is cast in the form of learning an end-to-end mapping from observed data to the ground-truth. Inspired by the splitting strategy widely used in regularized iterative algorithm to tackle inverse problems, the mapping is decomposed into two networks, with one handling the inversion of the physical forward model associated with the data term and one handling the denoising of the output from the former network, i.e., the inverted version, associated with the prior/regularization term. The two networks are trained jointly to learn the end-to-end mapping, getting rid of a two-step training. The training is annealing as the intermediate variable between these two networks bridges the gap between the input (the degraded version of output) and output and progressively approaches to the ground-truth. The proposed network, referred to as InverseNet, is flexible in the sense that most of the existing end-to-end network structure can be leveraged in the first network and most of the existing denoising network structure can be used in the second one. Extensive experiments on both synthetic data and real datasets on the tasks, motion deblurring, super-resolution, and colorization, demonstrate the efficiency and accuracy of the proposed method compared with other image processing algorithms.

Wenlin Wang, Yunchen Pu, Vinay Kumar Verma et al.

We present a deep generative model for learning to predict classes not seen at training time. Unlike most existing methods for this problem, that represent each class as a point (via a semantic embedding), we represent each seen/unseen class using a class-specific latent-space distribution, conditioned on class attributes. We use these latent-space distributions as a prior for a supervised variational autoencoder (VAE), which also facilitates learning highly discriminative feature representations for the inputs. The entire framework is learned end-to-end using only the seen-class training data. The model infers corresponding attributes of a test image by maximizing the VAE lower bound; the inferred attributes may be linked to labels not seen when training. We further extend our model to a (1) semi-supervised/transductive setting by leveraging unlabeled unseen-class data via an unsupervised learning module, and (2) few-shot learning where we also have a small number of labeled inputs from the unseen classes. We compare our model with several state-of-the-art methods through a comprehensive set of experiments on a variety of benchmark data sets.

Changyou Chen, Wenlin Wang, Yizhe Zhang et al.

Stochastic gradient Markov Chain Monte Carlo (SG-MCMC) has been developed as a flexible family of scalable Bayesian sampling algorithms. However, there has been little theoretical analysis of the impact of minibatch size to the algorithm's convergence rate. In this paper, we prove that under a limited computational budget/time, a larger minibatch size leads to a faster decrease of the mean squared error bound (thus the fastest one corresponds to using full gradients), which motivates the necessity of variance reduction in SG-MCMC. Consequently, by borrowing ideas from stochastic optimization, we propose a practical variance-reduction technique for SG-MCMC, that is efficient in both computation and storage. We develop theory to prove that our algorithm induces a faster convergence rate than standard SG-MCMC. A number of large-scale experiments, ranging from Bayesian learning of logistic regression to deep neural networks, validate the theory and demonstrate the superiority of the proposed variance-reduction SG-MCMC framework.

Changyou Chen, Chunyuan Li, Liqun Chen et al.

Two fundamental problems in unsupervised learning are efficient inference for latent-variable models and robust density estimation based on large amounts of unlabeled data. Algorithms for the two tasks, such as normalizing flows and generative adversarial networks (GANs), are often developed independently. In this paper, we propose the concept of {\em continuous-time flows} (CTFs), a family of diffusion-based methods that are able to asymptotically approach a target distribution. Distinct from normalizing flows and GANs, CTFs can be adopted to achieve the above two goals in one framework, with theoretical guarantees. Our framework includes distilling knowledge from a CTF for efficient inference, and learning an explicit energy-based distribution with CTFs for density estimation. Both tasks rely on a new technique for distribution matching within amortized learning. Experiments on various tasks demonstrate promising performance of the proposed CTF framework, compared to related techniques.

Wenlin Wang, Changyou Chen, Wenqi Wang et al.

We present Earliness-Aware Deep Convolutional Networks (EA-ConvNets), an end-to-end deep learning framework, for early classification of time series data. Unlike most existing methods for early classification of time series data, that are designed to solve this problem under the assumption of the availability of a good set of pre-defined (often hand-crafted) features, our framework can jointly perform feature learning (by learning a deep hierarchy of \emph{shapelets} capturing the salient characteristics in each time series), along with a dynamic truncation model to help our deep feature learning architecture focus on the early parts of each time series. Consequently, our framework is able to make highly reliable early predictions, outperforming various state-of-the-art methods for early time series classification, while also being competitive when compared to the state-of-the-art time series classification algorithms that work with \emph{fully observed} time series data. To the best of our knowledge, the proposed framework is the first to perform data-driven (deep) feature learning in the context of early classification of time series data. We perform a comprehensive set of experiments, on several benchmark data sets, which demonstrate that our method yields significantly better predictions than various state-of-the-art methods designed for early time series classification. In addition to obtaining high accuracies, our experiments also show that the learned deep shapelets based features are also highly interpretable and can help gain better understanding of the underlying characteristics of time series data.