Weiran Wang

Zhiheng Xi, Wenxiang Chen, Xin Guo et al.

For a long time, humanity has pursued artificial intelligence (AI) equivalent to or surpassing the human level, with AI agents considered a promising vehicle for this pursuit. AI agents are artificial entities that sense their environment, make decisions, and take actions. Many efforts have been made to develop intelligent agents, but they mainly focus on advancement in algorithms or training strategies to enhance specific capabilities or performance on particular tasks. Actually, what the community lacks is a general and powerful model to serve as a starting point for designing AI agents that can adapt to diverse scenarios. Due to the versatile capabilities they demonstrate, large language models (LLMs) are regarded as potential sparks for Artificial General Intelligence (AGI), offering hope for building general AI agents. Many researchers have leveraged LLMs as the foundation to build AI agents and have achieved significant progress. In this paper, we perform a comprehensive survey on LLM-based agents. We start by tracing the concept of agents from its philosophical origins to its development in AI, and explain why LLMs are suitable foundations for agents. Building upon this, we present a general framework for LLM-based agents, comprising three main components: brain, perception, and action, and the framework can be tailored for different applications. Subsequently, we explore the extensive applications of LLM-based agents in three aspects: single-agent scenarios, multi-agent scenarios, and human-agent cooperation. Following this, we delve into agent societies, exploring the behavior and personality of LLM-based agents, the social phenomena that emerge from an agent society, and the insights they offer for human society. Finally, we discuss several key topics and open problems within the field. A repository for the related papers at https://github.com/WooooDyy/LLM-Agent-Paper-List.

Yawei Li, Kai Zhang, Radu Timofte et al. · eth-zurich, tencent-ai

This paper reviews the NTIRE 2022 challenge on efficient single image super-resolution with focus on the proposed solutions and results. The task of the challenge was to super-resolve an input image with a magnification factor of $\times$4 based on pairs of low and corresponding high resolution images. The aim was to design a network for single image super-resolution that achieved improvement of efficiency measured according to several metrics including runtime, parameters, FLOPs, activations, and memory consumption while at least maintaining the PSNR of 29.00dB on DIV2K validation set. IMDN is set as the baseline for efficiency measurement. The challenge had 3 tracks including the main track (runtime), sub-track one (model complexity), and sub-track two (overall performance). In the main track, the practical runtime performance of the submissions was evaluated. The rank of the teams were determined directly by the absolute value of the average runtime on the validation set and test set. In sub-track one, the number of parameters and FLOPs were considered. And the individual rankings of the two metrics were summed up to determine a final ranking in this track. In sub-track two, all of the five metrics mentioned in the description of the challenge including runtime, parameter count, FLOPs, activations, and memory consumption were considered. Similar to sub-track one, the rankings of five metrics were summed up to determine a final ranking. The challenge had 303 registered participants, and 43 teams made valid submissions. They gauge the state-of-the-art in efficient single image super-resolution.

Tara N. Sainath, Rohit Prabhavalkar, Ankur Bapna et al.

We present JOIST, an algorithm to train a streaming, cascaded, encoder end-to-end (E2E) model with both speech-text paired inputs, and text-only unpaired inputs. Unlike previous works, we explore joint training with both modalities, rather than pre-training and fine-tuning. In addition, we explore JOIST using a streaming E2E model with an order of magnitude more data, which are also novelties compared to previous works. Through a series of ablation studies, we explore different types of text modeling, including how to model the length of the text sequence and the appropriate text sub-word unit representation. We find that best text representation for JOIST improves WER across a variety of search and rare-word test sets by 4-14% relative, compared to a model not trained with text. In addition, we quantitatively show that JOIST maintains streaming capabilities, which is important for good user-level experience.

Shaojin Ding, Weiran Wang, Ding Zhao et al.

In this paper, we propose a dynamic cascaded encoder Automatic Speech Recognition (ASR) model, which unifies models for different deployment scenarios. Moreover, the model can significantly reduce model size and power consumption without loss of quality. Namely, with the dynamic cascaded encoder model, we explore three techniques to maximally boost the performance of each model size: 1) Use separate decoders for each sub-model while sharing the encoders; 2) Use funnel-pooling to improve the encoder efficiency; 3) Balance the size of causal and non-causal encoders to improve quality and fit deployment constraints. Overall, the proposed large-medium model has 30% smaller size and reduces power consumption by 33%, compared to the baseline cascaded encoder model. The triple-size model that unifies the large, medium, and small models achieves 37% total size reduction with minimal quality loss, while substantially reducing the engineering efforts of having separate models.

Zhong Meng, Weiran Wang, Rohit Prabhavalkar et al.

We propose JEIT, a joint end-to-end (E2E) model and internal language model (ILM) training method to inject large-scale unpaired text into ILM during E2E training which improves rare-word speech recognition. With JEIT, the E2E model computes an E2E loss on audio-transcript pairs while its ILM estimates a cross-entropy loss on unpaired text. The E2E model is trained to minimize a weighted sum of E2E and ILM losses. During JEIT, ILM absorbs knowledge from unpaired text while the E2E training serves as regularization. Unlike ILM adaptation methods, JEIT does not require a separate adaptation step and avoids the need for Kullback-Leibler divergence regularization of ILM. We also show that modular hybrid autoregressive transducer (MHAT) performs better than HAT in the JEIT framework, and is much more robust than HAT during ILM adaptation. To push the limit of unpaired text injection, we further propose a combined JEIT and JOIST training (CJJT) that benefits from modality matching, encoder text injection and ILM training. Both JEIT and CJJT can foster a more effective LM fusion. With 100B unpaired sentences, JEIT/CJJT improves rare-word recognition accuracy by up to 16.4% over a model trained without unpaired text.

Weiran Wang, Tongzhou Chen, Tara N. Sainath et al.

Language models (LMs) significantly improve the recognition accuracy of end-to-end (E2E) models on words rarely seen during training, when used in either the shallow fusion or the rescoring setups. In this work, we introduce LMs in the learning of hybrid autoregressive transducer (HAT) models in the discriminative training framework, to mitigate the training versus inference gap regarding the use of LMs. For the shallow fusion setup, we use LMs during both hypotheses generation and loss computation, and the LM-aware MWER-trained model achieves 10\% relative improvement over the model trained with standard MWER on voice search test sets containing rare words. For the rescoring setup, we learn a small neural module to generate per-token fusion weights in a data-dependent manner. This model achieves the same rescoring WER as regular MWER-trained model, but without the need for sweeping fusion weights.

Ke Hu, Tara N. Sainath, Yanzhang He et al.

Text-only and semi-supervised training based on audio-only data has gained popularity recently due to the wide availability of unlabeled text and speech data. In this work, we propose incorporating text-only and semi-supervised training into an attention-based deliberation model. By incorporating text-only data in training a bidirectional encoder representation from transformer (BERT) for the deliberation text encoder, and large-scale text-to-speech and audio-only utterances using joint acoustic and text decoder (JATD) and semi-supervised training, we achieved 4%-12% WER reduction for various tasks compared to the baseline deliberation. Compared to a state-of-the-art language model (LM) rescoring method, the deliberation model reduces the Google Voice Search WER by 11% relative. We show that the deliberation model also achieves a positive human side-by-side evaluation compared to the state-of-the-art LM rescorer with reasonable endpointer latencies.

Shaan Bijwadia, Shuo-yiin Chang, Weiran Wang et al.

Text injection for automatic speech recognition (ASR), wherein unpaired text-only data is used to supplement paired audio-text data, has shown promising improvements for word error rate. This study examines the use of text injection for auxiliary tasks, which are the non-ASR tasks often performed by an E2E model. In this work, we use joint end-to-end and internal language model training (JEIT) as our text injection algorithm to train an ASR model which performs two auxiliary tasks. The first is capitalization, which is a de-normalization task. The second is turn-taking prediction, which attempts to identify whether a user has completed their conversation turn in a digital assistant interaction. We show results demonstrating that our text injection method boosts capitalization performance for long-tail data, and improves turn-taking detection recall.

Rami Botros, Anmol Gulati, Tara N. Sainath et al.

Conformer models maintain a large number of internal states, the vast majority of which are associated with self-attention layers. With limited memory bandwidth, reading these from memory at each inference step can slow down inference. In this paper, we design an optimized conformer that is small enough to meet on-device restrictions and has fast inference on TPUs. We explore various ideas to improve the execution speed, including replacing lower conformer blocks with convolution-only blocks, strategically downsizing the architecture, and utilizing an RNNAttention-Performer. Our optimized conformer can be readily incorporated into a cascaded-encoder setting, allowing a second-pass decoder to operate on its output and improve the accuracy whenever more resources are available. Altogether, we find that these optimizations can reduce latency by a factor of 6.8x, and come at a reasonable trade-off in quality. With the cascaded second-pass, we show that the recognition accuracy is completely recoverable. Thus, our proposed encoder can double as a strong standalone encoder in on device, and as the first part of a high-performance ASR pipeline.

Weiran Wang, Ke Hu, Tara N. Sainath

We propose a streaming non-autoregressive (non-AR) decoding algorithm to deliberate the hypothesis alignment of a streaming RNN-T model. Our algorithm facilitates a simple greedy decoding procedure, and at the same time is capable of producing the decoding result at each frame with limited right context, thus enjoying both high efficiency and low latency. These advantages are achieved by converting the offline Align-Refine algorithm to be streaming-compatible, with a novel transformer decoder architecture that performs local self-attentions for both text and audio, and a time-aligned cross-attention at each layer. Furthermore, we perform discriminative training of our model with the minimum word error rate (MWER) criterion, which has not been done in the non-AR decoding literature. Experiments on voice search datasets and Librispeech show that with reasonable right context, our streaming model performs as well as the offline counterpart, and discriminative training leads to further WER gain when the first-pass model has small capacity.

Weiran Wang, Zelin Wu, Diamantino Caseiro et al.

Contextual biasing refers to the problem of biasing the automatic speech recognition (ASR) systems towards rare entities that are relevant to the specific user or application scenarios. We propose algorithms for contextual biasing based on the Knuth-Morris-Pratt algorithm for pattern matching. During beam search, we boost the score of a token extension if it extends matching into a set of biasing phrases. Our method simulates the classical approaches often implemented in the weighted finite state transducer (WFST) framework, but avoids the FST language altogether, with careful considerations on memory footprint and efficiency on tensor processing units (TPUs) by vectorization. Without introducing additional model parameters, our method achieves significant word error rate (WER) reductions on biasing test sets by itself, and yields further performance gain when combined with a model-based biasing method.

Yiling Huang, Weiran Wang, Guanlong Zhao et al.

While standard speaker diarization attempts to answer the question "who spoken when", most of relevant applications in reality are more interested in determining "who spoken what". Whether it is the conventional modularized approach or the more recent end-to-end neural diarization (EEND), an additional automatic speech recognition (ASR) model and an orchestration algorithm are required to associate the speaker labels with recognized words. In this paper, we propose Word-level End-to-End Neural Diarization (WEEND) with auxiliary network, a multi-task learning algorithm that performs end-to-end ASR and speaker diarization in the same neural architecture. That is, while speech is being recognized, speaker labels are predicted simultaneously for each recognized word. Experimental results demonstrate that WEEND outperforms the turn-based diarization baseline system on all 2-speaker short-form scenarios and has the capability to generalize to audio lengths of 5 minutes. Although 3+speaker conversations are harder, we find that with enough in-domain training data, WEEND has the potential to deliver high quality diarized text.

Yiyang Shen, Lifu Tu, Weiran Wang

Reinforcement Learning (RL) has been shown to substantially improve the reasoning capability of small and large language models (LLMs), but existing approaches typically rely on verifiable rewards, hence ground truth labels. We propose an RL framework that uses rewards from an LLM that acts as a judge evaluating model outputs over large amounts of unlabeled data, enabling label-free knowledge distillation and replacing the need of ground truth supervision. Notably, the judge operates with a single-token output, making reward computation efficient. When combined with verifiable rewards, our approach yields substantial performance gains across math reasoning benchmarks. These results suggest that LLM-based evaluators can produce effective training signals for RL fine-tuning.

Haozhe Shan, Albert Gu, Zhong Meng et al.

Online speech recognition, where the model only accesses context to the left, is an important and challenging use case for ASR systems. In this work, we investigate augmenting neural encoders for online ASR by incorporating structured state-space sequence models (S4), a family of models that provide a parameter-efficient way of accessing arbitrarily long left context. We performed systematic ablation studies to compare variants of S4 models and propose two novel approaches that combine them with convolutions. We found that the most effective design is to stack a small S4 using real-valued recurrent weights with a local convolution, allowing them to work complementarily. Our best model achieves WERs of 4.01%/8.53% on test sets from Librispeech, outperforming Conformers with extensively tuned convolution.

Yiyang Shen, Yutian He, Weiran Wang et al.

We study a class of bilevel optimization problems in which both the upper- and lower-level problems have minimax structures. This setting captures a broad range of emerging applications. Despite the extensive literature on bilevel optimization and minimax optimization separately, existing methods mainly focus on bilevel optimization with lower-level minimization problems, often under strong convexity assumptions, and are not directly applicable to the minimax lower-level setting considered here. To address this gap, we develop penalty-based first-order methods for bilevel minimax optimization without requiring strong convexity of the lower-level problem. In the deterministic setting, we establish that the proposed method finds an $ε$-KKT point with $\tilde{O}(ε^{-4})$ oracle complexity. We further show that bilevel problems with convex constrained lower-level minimization can be reformulated as special cases of our framework via Lagrangian duality, leading to an $\tilde{O}(ε^{-4})$ complexity bound that improves upon the existing $\tilde{O}(ε^{-7})$ result. Finally, we extend our approach to the stochastic setting, where only stochastic gradient oracles are available, and prove that the proposed stochastic method finds a nearly $ε$-KKT point with $\tilde{O}(ε^{-9})$ oracle complexity.

Dongseong Hwang, Weiran Wang, Zhuoyuan Huo et al.

While Transformers have revolutionized deep learning, their quadratic attention complexity hinders their ability to process infinitely long inputs. We propose Feedback Attention Memory (FAM), a novel Transformer architecture that leverages a feedback loop to enable the network to attend to its own latent representations. This design fosters the emergence of working memory within the Transformer, allowing it to process indefinitely long sequences. TransformerFAM requires no additional weights, enabling seamless integration with pre-trained models. Our experiments show that TransformerFAM significantly improves Transformer performance on long-context tasks across various model sizes (1B, 8B, and 24B). These results showcase the potential to empower Large Language Models (LLMs) to process sequences of unlimited length.

Rohit Prabhavalkar, Zhong Meng, Weiran Wang et al.

The accuracy of end-to-end (E2E) automatic speech recognition (ASR) models continues to improve as they are scaled to larger sizes, with some now reaching billions of parameters. Widespread deployment and adoption of these models, however, requires computationally efficient strategies for decoding. In the present work, we study one such strategy: applying multiple frame reduction layers in the encoder to compress encoder outputs into a small number of output frames. While similar techniques have been investigated in previous work, we achieve dramatically more reduction than has previously been demonstrated through the use of multiple funnel reduction layers. Through ablations, we study the impact of various architectural choices in the encoder to identify the most effective strategies. We demonstrate that we can generate one encoder output frame for every 2.56 sec of input speech, without significantly affecting word error rate on a large-scale voice search task, while improving encoder and decoder latencies by 48% and 92% respectively, relative to a strong but computationally expensive baseline.

Zelin Wu, Gan Song, Christopher Li et al.

Contextual biasing enables speech recognizers to transcribe important phrases in the speaker's context, such as contact names, even if they are rare in, or absent from, the training data. Attention-based biasing is a leading approach which allows for full end-to-end cotraining of the recognizer and biasing system and requires no separate inference-time components. Such biasers typically consist of a context encoder; followed by a context filter which narrows down the context to apply, improving per-step inference time; and, finally, context application via cross attention. Though much work has gone into optimizing per-frame performance, the context encoder is at least as important: recognition cannot begin before context encoding ends. Here, we show the lightweight phrase selection pass can be moved before context encoding, resulting in a speedup of up to 16.1 times and enabling biasing to scale to 20K phrases with a maximum pre-decoding delay under 33ms. With the addition of phrase- and wordpiece-level cross-entropy losses, our technique also achieves up to a 37.5% relative WER reduction over the baseline without the losses and lightweight phrase selection pass.

Yitong Duan, Weiran Wang, Jian Li

As a fundamental method in economics and finance, the factor model has been extensively utilized in quantitative investment. In recent years, there has been a paradigm shift from traditional linear models with expert-designed factors to more flexible nonlinear machine learning-based models with data-driven factors, aiming to enhance the effectiveness of these factor models. However, due to the low signal-to-noise ratio in market data, mining effective factors in data-driven models remains challenging. In this work, we propose a hypergraph-based factor model with temporal residual contrastive learning (FactorGCL) that employs a hypergraph structure to better capture high-order nonlinear relationships among stock returns and factors. To mine hidden factors that supplement human-designed prior factors for predicting stock returns, we design a cascading residual hypergraph architecture, in which the hidden factors are extracted from the residual information after removing the influence of prior factors. Additionally, we propose a temporal residual contrastive learning method to guide the extraction of effective and comprehensive hidden factors by contrasting stock-specific residual information over different time periods. Our extensive experiments on real stock market data demonstrate that FactorGCL not only outperforms existing state-of-the-art methods but also mines effective hidden factors for predicting stock returns.

Wenjuan Qin, Weiran Wang, Yuming Yang et al.

The study investigates the efficacy of pre-trained language models (PLMs) in analyzing argumentative moves in a longitudinal learner corpus. Prior studies on argumentative moves often rely on qualitative analysis and manual coding, limiting their efficiency and generalizability. The study aims to: 1) to assess the reliability of PLMs in analyzing argumentative moves; 2) to utilize PLM-generated annotations to illustrate developmental patterns and predict writing quality. A longitudinal corpus of 1643 argumentative texts from 235 English learners in China is collected and annotated into six move types: claim, data, counter-claim, counter-data, rebuttal, and non-argument. The corpus is divided into training, validation, and application sets annotated by human experts and PLMs. We use BERT as one of the implementations of PLMs. The results indicate a robust reliability of PLMs in analyzing argumentative moves, with an overall F1 score of 0.743, surpassing existing models in the field. Additionally, PLM-labeled argumentative moves effectively capture developmental patterns and predict writing quality. Over time, students exhibit an increase in the use of data and counter-claims and a decrease in non-argument moves. While low-quality texts are characterized by a predominant use of claims and data supporting only oneside position, mid- and high-quality texts demonstrate an integrative perspective with a higher ratio of counter-claims, counter-data, and rebuttals. This study underscores the transformative potential of integrating artificial intelligence into language education, enhancing the efficiency and accuracy of evaluating students' writing. The successful application of PLMs can catalyze the development of educational technology, promoting a more data-driven and personalized learning environment that supports diverse educational needs.

Wanting Huang, Weiran Wang

Contextual biasing improves automatic speech recognition (ASR) by integrating external knowledge, such as user-specific phrases or entities, during decoding. In this work, we use an attention-based biasing decoder to produce scores for candidate phrases based on acoustic information extracted by an ASR encoder, which can be used to filter out unlikely phrases and to calculate bonus for shallow-fusion biasing. We introduce a per-token discriminative objective that encourages higher scores for ground-truth phrases while suppressing distractors. Experiments on the Librispeech biasing benchmark show that our method effectively filters out majority of the candidate phrases, and significantly improves recognition accuracy under different biasing conditions when the scores are used in shallow fusion biasing. Our approach is modular and can be used with any ASR system, and the filtering mechanism can potentially boost performance of other biasing methods.

Yijie Zhang, Yiyang Shen, Weiran Wang

Multimodal data are prevalent across various domains, and learning robust representations of such data is paramount to enhancing generation quality and downstream task performance. To handle heterogeneity and interconnections among different modalities, recent multimodal generative models extract shared and private (modality-specific) information with two separate variables. Despite attempts to enforce disentanglement between these two variables, these methods struggle with challenging datasets where the likelihood model is insufficient. In this paper, we propose Information-disentangled Multimodal VAE (IDMVAE) to explicitly address this issue, with rigorous mutual information-based regularizations, including cross-view mutual information maximization for extracting shared variables, and a cycle-consistency style loss for redundancy removal using generative augmentations. We further introduce diffusion models to improve the capacity of latent priors. These newly proposed components are complementary to each other. Compared to existing approaches, IDMVAE shows a clean separation between shared and private information, demonstrating superior generation quality and semantic coherence on challenging datasets.

Shanmuka Sadhu, Weiran Wang

Consistency regularization (CR), which enforces agreement between model predictions on augmented views, has found recent benefits in automatic speech recognition [1]. In this paper, we propose the use of consistency regularization for audio event recognition, and demonstrate its effectiveness on AudioSet. With extensive ablation studies for both small ($\sim$20k) and large ($\sim$1.8M) supervised training sets, we show that CR brings consistent improvement over supervised baselines which already heavily utilize data augmentation, and CR using stronger augmentation and multiple augmentations leads to additional gain for the small training set. Furthermore, we extend the use of CR into the semi-supervised setup with 20K labeled samples and 1.8M unlabeled samples, and obtain performance improvement over our best model trained on the small set.

Yiyang Shen, Weiran Wang

Multi-view learning is widely applied to real-life datasets, such as multiple omics biological data, but it often suffers from both missing views and missing labels. Prior probabilistic approaches addressed the missing view problem by using a product-of-experts scheme to aggregate representations from present views and achieved superior performance over deterministic classifiers, using the information bottleneck (IB) principle. However, the IB framework is inherently fully supervised and cannot leverage unlabeled data. In this work, we propose a semi-supervised generative model that utilizes both labeled and unlabeled samples in a unified framework. Our method maximizes the likelihood of unlabeled samples to learn a latent space shared with the IB on labeled data. We also perform cross-view mutual information maximization in the latent space to enhance the extraction of shared information across views. Compared to existing approaches, our model achieves better predictive and imputation performance on both image and multi-omics data with missing views and limited labeled samples.

Li Xu, Siqi Wang, Kepeng Xu et al.

High-Dynamic-Range Wide-Color-Gamut (HDR-WCG) technology is becoming increasingly widespread, driving a growing need for converting Standard Dynamic Range (SDR) content to HDR. Existing methods primarily rely on fixed tone mapping operators, which struggle to handle the diverse appearances and degradations commonly present in real-world SDR content. To address this limitation, we propose a generalized SDR-to-HDR framework that enhances robustness by learning attribute-disentangled representations. Central to our approach is Realistic Attribute-Disentangled Representation Learning (RealRep), which explicitly disentangles luminance and chrominance components to capture intrinsic content variations across different SDR distributions. Furthermore, we design a Luma-/Chroma-aware negative exemplar generation strategy that constructs degradation-sensitive contrastive pairs, effectively modeling tone discrepancies across SDR styles. Building on these attribute-level priors, we introduce the Degradation-Domain Aware Controlled Mapping Network (DDACMNet), a lightweight, two-stage framework that performs adaptive hierarchical mapping guided by a control-aware normalization mechanism. DDACMNet dynamically modulates the mapping process via degradation-conditioned features, enabling robust adaptation across diverse degradation domains. Extensive experiments demonstrate that RealRep consistently outperforms state-of-the-art methods in both generalization and perceptually faithful HDR color gamut reconstruction.

Zhong Meng, Zelin Wu, Rohit Prabhavalkar et al.

Neural contextual biasing effectively improves automatic speech recognition (ASR) for crucial phrases within a speaker's context, particularly those that are infrequent in the training data. This work proposes contextual text injection (CTI) to enhance contextual ASR. CTI leverages not only the paired speech-text data, but also a much larger corpus of unpaired text to optimize the ASR model and its biasing component. Unpaired text is converted into speech-like representations and used to guide the model's attention towards relevant bias phrases. Moreover, we introduce a contextual text-injected (CTI) minimum word error rate (MWER) training, which minimizes the expected WER caused by contextual biasing when unpaired text is injected into the model. Experiments show that CTI with 100 billion text sentences can achieve up to 43.3% relative WER reduction from a strong neural biasing model. CTI-MWER provides a further relative improvement of 23.5%.

Weiran Wang, Ke Hu, Tara Sainath

We propose to deliberate the hypothesis alignment of a streaming RNN-T model with the previously proposed Align-Refine non-autoregressive decoding method and its improved versions. The method performs a few refinement steps, where each step shares a transformer decoder that attends to both text features (extracted from alignments) and audio features, and outputs complete updated alignments. The transformer decoder is trained with the CTC loss which facilitates parallel greedy decoding, and performs full-context attention to capture label dependencies. We improve Align-Refine by introducing cascaded encoder that captures more audio context before refinement, and alignment augmentation which enforces learning label dependency. We show that, conditioned on hypothesis alignments of a streaming RNN-T model, our method obtains significantly more accurate recognition results than the first-pass RNN-T, with only small amount of model parameters.

Junwen Bai, Weiran Wang, Carla Gomes

Self-supervised disentangled representation learning is a critical task in sequence modeling. The learnt representations contribute to better model interpretability as well as the data generation, and improve the sample efficiency for downstream tasks. We propose a novel sequence representation learning method, named Contrastively Disentangled Sequential Variational Autoencoder (C-DSVAE), to extract and separate the static (time-invariant) and dynamic (time-variant) factors in the latent space. Different from previous sequential variational autoencoder methods, we use a novel evidence lower bound which maximizes the mutual information between the input and the latent factors, while penalizes the mutual information between the static and dynamic factors. We leverage contrastive estimations of the mutual information terms in training, together with simple yet effective augmentation techniques, to introduce additional inductive biases. Our experiments show that C-DSVAE significantly outperforms the previous state-of-the-art methods on multiple metrics.

Qi Lyu, Xiao Fu, Weiran Wang et al.

Multiple views of data, both naturally acquired (e.g., image and audio) and artificially produced (e.g., via adding different noise to data samples), have proven useful in enhancing representation learning. Natural views are often handled by multiview analysis tools, e.g., (deep) canonical correlation analysis [(D)CCA], while the artificial ones are frequently used in self-supervised learning (SSL) paradigms, e.g., BYOL and Barlow Twins. Both types of approaches often involve learning neural feature extractors such that the embeddings of data exhibit high cross-view correlations. Although intuitive, the effectiveness of correlation-based neural embedding is mostly empirically validated. This work aims to understand latent correlation maximization-based deep multiview learning from a latent component identification viewpoint. An intuitive generative model of multiview data is adopted, where the views are different nonlinear mixtures of shared and private components. Since the shared components are view/distortion-invariant, representing the data using such components is believed to reveal the identity of the samples effectively and robustly. Under this model, latent correlation maximization is shown to guarantee the extraction of the shared components across views (up to certain ambiguities). In addition, it is further shown that the private information in each view can be provably disentangled from the shared using proper regularization design. A finite sample analysis, which has been rare in nonlinear mixture identifiability study, is also presented. The theoretical results and newly designed regularization are tested on a series of tasks.

Junwen Bai, Weiran Wang, Yingbo Zhou et al.

We propose Deep Autoencoding Predictive Components (DAPC) -- a self-supervised representation learning method for sequence data, based on the intuition that useful representations of sequence data should exhibit a simple structure in the latent space. We encourage this latent structure by maximizing an estimate of predictive information of latent feature sequences, which is the mutual information between past and future windows at each time step. In contrast to the mutual information lower bound commonly used by contrastive learning, the estimate of predictive information we adopt is exact under a Gaussian assumption. Additionally, it can be computed without negative sampling. To reduce the degeneracy of the latent space extracted by powerful encoders and keep useful information from the inputs, we regularize predictive information learning with a challenging masked reconstruction loss. We demonstrate that our method recovers the latent space of noisy dynamical systems, extracts predictive features for forecasting tasks, and improves automatic speech recognition when used to pretrain the encoder on large amounts of unlabeled data.

Chieh-Chi Kao, Ming Sun, Weiran Wang et al.

Acoustic event classification (AEC) and acoustic event detection (AED) refer to the task of detecting whether specific target events occur in audios. As long short-term memory (LSTM) leads to state-of-the-art results in various speech related tasks, it is employed as a popular solution for AEC as well. This paper focuses on investigating the dynamics of LSTM model on AEC tasks. It includes a detailed analysis on LSTM memory retaining, and a benchmarking of nine different pooling methods on LSTM models using 1.7M generated mixture clips of multiple events with different signal-to-noise ratios. This paper focuses on understanding: 1) utterance-level classification accuracy; 2) sensitivity to event position within an utterance. The analysis is done on the dataset for the detection of rare sound events from DCASE 2017 Challenge. We find max pooling on the prediction level to perform the best among the nine pooling approaches in terms of classification accuracy and insensitivity to event position within an utterance. To authors' best knowledge, this is the first kind of such work focused on LSTM dynamics for AEC tasks.

Weiran Wang, Qingming Tang, Karen Livescu

We propose an approach for pre-training speech representations via a masked reconstruction loss. Our pre-trained encoder networks are bidirectional and can therefore be used directly in typical bidirectional speech recognition models. The pre-trained networks can then be fine-tuned on a smaller amount of supervised data for speech recognition. Experiments with this approach on the LibriSpeech and Wall Street Journal corpora show promising results. We find that the main factors that lead to speech recognition improvements are: masking segments of sufficient width in both time and frequency, pre-training on a much larger amount of unlabeled data than the labeled data, and domain adaptation when the unlabeled and labeled data come from different domains. The gain from pre-training is additive to that of supervised data augmentation.

Yang Chen, Weiran Wang, I-Fan Chen et al.

Practitioners often need to build ASR systems for new use cases in a short amount of time, given limited in-domain data. While recently developed end-to-end methods largely simplify the modeling pipelines, they still suffer from the data sparsity issue. In this work, we explore a few simple-to-implement techniques for building online ASR systems in an end-to-end fashion, with a small amount of transcribed data in the target domain. These techniques include data augmentation in the target domain, domain adaptation using models previously trained on a large source domain, and knowledge distillation on non-transcribed target domain data, using an adapted bi-directional model as the teacher; they are applicable in real scenarios with different types of resources. Our experiments demonstrate that each technique is independently useful in the improvement of the online ASR performance in the target domain.

Yang Chen, Weiran Wang, Chao Wang

While deep learning based end-to-end automatic speech recognition (ASR) systems have greatly simplified modeling pipelines, they suffer from the data sparsity issue. In this work, we propose a self-training method with an end-to-end system for semi-supervised ASR. Starting from a Connectionist Temporal Classification (CTC) system trained on the supervised data, we iteratively generate pseudo-labels on a mini-batch of unsupervised utterances with the current model, and use the pseudo-labels to augment the supervised data for immediate model update. Our method retains the simplicity of end-to-end ASR systems, and can be seen as performing alternating optimization over a well-defined learning objective. We also perform empirical investigations of our method, regarding the effect of data augmentation, decoding beamsize for pseudo-label generation, and freshness of pseudo-labels. On a commonly used semi-supervised ASR setting with the WSJ corpus, our method gives 14.4% relative WER improvement over a carefully-trained base system with data augmentation, reducing the performance gap between the base system and the oracle system by 50%.

Weimin Wang, Weiran Wang, Ming Sun et al.

Acoustic Scene Classification (ASC) is a challenging task, as a single scene may involve multiple events that contain complex sound patterns. For example, a cooking scene may contain several sound sources including silverware clinking, chopping, frying, etc. What complicates ASC more is that classes of different activities could have overlapping sounds patterns (e.g. both cooking and dishwashing could have silverware clinking sound). In this paper, we propose a multi-head attention network to model the complex temporal input structures for ASC. The proposed network takes the audio's time-frequency representation as input, and it leverages standard VGG plus LSTM layers to extract high-level feature representation. Further more, it applies multiple attention heads to summarize various patterns of sound events into fixed dimensional representation, for the purpose of final scene classification. The whole network is trained in an end-to-end fashion with back-propagation. Experimental results confirm that our model discovers meaningful sound patterns through the attention mechanism, without using explicit supervision in the alignment. We evaluated our proposed model using DCASE 2018 Task 5 dataset, and achieved competitive performance on par with previous winner's results.

Gustavo Aguilar, Viktor Rozgić, Weiran Wang et al.

Studies on emotion recognition (ER) show that combining lexical and acoustic information results in more robust and accurate models. The majority of the studies focus on settings where both modalities are available in training and evaluation. However, in practice, this is not always the case; getting ASR output may represent a bottleneck in a deployment pipeline due to computational complexity or privacy-related constraints. To address this challenge, we study the problem of efficiently combining acoustic and lexical modalities during training while still providing a deployable acoustic model that does not require lexical inputs. We first experiment with multimodal models and two attention mechanisms to assess the extent of the benefits that lexical information can provide. Then, we frame the task as a multi-view learning problem to induce semantic information from a multimodal model into our acoustic-only network using a contrastive loss function. Our multimodal model outperforms the previous state of the art on the USC-IEMOCAP dataset reported on lexical and acoustic information. Additionally, our multi-view-trained acoustic network significantly surpasses models that have been exclusively trained with acoustic features.

Weiran Wang

We adopt a multi-view approach for analyzing two knowledge transfer settings---learning using privileged information (LUPI) and distillation---in a common framework. Under reasonable assumptions about the complexities of hypothesis spaces, and being optimistic about the expected loss achievable by the student (in distillation) and a transformed teacher predictor (in LUPI), we show that encouraging agreement between the teacher and the student leads to reduced search space. As a result, improved convergence rate can be obtained with regularized empirical risk minimization.

Weiran Wang, Chieh-chi Kao, Chao Wang

We propose a simple recurrent model for detecting rare sound events, when the time boundaries of events are available for training. Our model optimizes the combination of an utterance-level loss, which classifies whether an event occurs in an utterance, and a frame-level loss, which classifies whether each frame corresponds to the event when it does occur. The two losses make use of a shared vectorial representation the event, and are connected by an attention mechanism. We demonstrate our model on Task 2 of the DCASE 2017 challenge, and achieve competitive performance.

Chieh-Chi Kao, Weiran Wang, Ming Sun et al.

This paper proposes a Region-based Convolutional Recurrent Neural Network (R-CRNN) for audio event detection (AED). The proposed network is inspired by Faster-RCNN, a well known region-based convolutional network framework for visual object detection. Different from the original Faster-RCNN, a recurrent layer is added on top of the convolutional network to capture the long-term temporal context from the extracted high level features. While most of the previous works on AED generate predictions at frame level first, and then use post-processing to predict the onset/offset timestamps of events from a probability sequence; the proposed method generates predictions at event level directly and can be trained end-to-end with a multitask loss, which optimizes the classification and localization of audio events simultaneously. The proposed method is tested on DCASE 2017 Challenge dataset. To the best of our knowledge, R-CRNN is the best performing single-model method among all methods without using ensembles both on development and evaluation sets. Compared to the other region-based network for AED (R-FCN) with an event-based error rate (ER) of 0.18 on the development set, our method reduced the ER to half.

Qingming Tang, Weiran Wang, Karen Livescu

Previous work has shown that it is possible to improve speech recognition by learning acoustic features from paired acoustic-articulatory data, for example by using canonical correlation analysis (CCA) or its deep extensions. One limitation of this prior work is that the learned feature models are difficult to port to new datasets or domains, and articulatory data is not available for most speech corpora. In this work we study the problem of acoustic feature learning in the setting where we have access to an external, domain-mismatched dataset of paired speech and articulatory measurements, either with or without labels. We develop methods for acoustic feature learning in these settings, based on deep variational CCA and extensions that use both source and target domain data and labels. Using this approach, we improve phonetic recognition accuracies on both TIMIT and Wall Street Journal and analyze a number of design choices.

Weiran Wang, Jialei Wang, Mladen Kolar et al.

We propose methods for distributed graph-based multi-task learning that are based on weighted averaging of messages from other machines. Uniform averaging or diminishing stepsize in these methods would yield consensus (single task) learning. We show how simply skewing the averaging weights or controlling the stepsize allows learning different, but related, tasks on the different machines.

Weiran Wang, Nathan Srebro

We study stochastic optimization of nonconvex loss functions, which are typical objectives for training neural networks. We propose stochastic approximation algorithms which optimize a series of regularized, nonlinearized losses on large minibatches of samples, using only first-order gradient information. Our algorithms provably converge to an approximate critical point of the expected objective with faster rates than minibatch stochastic gradient descent, and facilitate better parallelization by allowing larger minibatches.

Qingming Tang, Weiran Wang, Karen Livescu

We study the problem of acoustic feature learning in the setting where we have access to another (non-acoustic) modality for feature learning but not at test time. We use deep variational canonical correlation analysis (VCCA), a recently proposed deep generative method for multi-view representation learning. We also extend VCCA with improved latent variable priors and with adversarial learning. Compared to other techniques for multi-view feature learning, VCCA's advantages include an intuitive latent variable interpretation and a variational lower bound objective that can be trained end-to-end efficiently. We compare VCCA and its extensions with previous feature learning methods on the University of Wisconsin X-ray Microbeam Database, and show that VCCA-based feature learning improves over previous methods for speaker-independent phonetic recognition.

Jialei Wang, Weiran Wang, Dan Garber et al.

We develop and analyze efficient "coordinate-wise" methods for finding the leading eigenvector, where each step involves only a vector-vector product. We establish global convergence with overall runtime guarantees that are at least as good as Lanczos's method and dominate it for slowly decaying spectrum. Our methods are based on combining a shift-and-invert approach with coordinate-wise algorithms for linear regression.

Jialei Wang, Weiran Wang, Nathan Srebro

We present and analyze an approach for distributed stochastic optimization which is statistically optimal and achieves near-linear speedups (up to logarithmic factors). Our approach allows a communication-memory tradeoff, with either logarithmic communication but linear memory, or polynomial communication and a corresponding polynomial reduction in required memory. This communication-memory tradeoff is achieved through minibatch-prox iterations (minibatch passive-aggressive updates), where a subproblem on a minibatch is solved at each iteration. We provide a novel analysis for such a minibatch-prox procedure which achieves the statistical optimal rate regardless of minibatch size and smoothness, thus significantly improving on prior work.

Chao Gao, Dan Garber, Nathan Srebro et al.

We study the sample complexity of canonical correlation analysis (CCA), \ie, the number of samples needed to estimate the population canonical correlation and directions up to arbitrarily small error. With mild assumptions on the data distribution, we show that in order to achieve $ε$-suboptimality in a properly defined measure of alignment between the estimated canonical directions and the population solution, we can solve the empirical objective exactly with $N(ε, Δ, γ)$ samples, where $Δ$ is the singular value gap of the whitened cross-covariance matrix and $1/γ$ is an upper bound of the condition number of auto-covariance matrices. Moreover, we can achieve the same learning accuracy by drawing the same level of samples and solving the empirical objective approximately with a stochastic optimization algorithm; this algorithm is based on the shift-and-invert power iterations and only needs to process the dataset for $\mathcal{O}\left(\log \frac{1}ε \right)$ passes. Finally, we show that, given an estimate of the canonical correlation, the streaming version of the shift-and-invert power iterations achieves the same learning accuracy with the same level of sample complexity, by processing the data only once.

Wanjia He, Weiran Wang, Karen Livescu

Recent work has begun exploring neural acoustic word embeddings---fixed-dimensional vector representations of arbitrary-length speech segments corresponding to words. Such embeddings are applicable to speech retrieval and recognition tasks, where reasoning about whole words may make it possible to avoid ambiguous sub-word representations. The main idea is to map acoustic sequences to fixed-dimensional vectors such that examples of the same word are mapped to similar vectors, while different-word examples are mapped to very different vectors. In this work we take a multi-view approach to learning acoustic word embeddings, in which we jointly learn to embed acoustic sequences and their corresponding character sequences. We use deep bidirectional LSTM embedding models and multi-view contrastive losses. We study the effect of different loss variants, including fixed-margin and cost-sensitive losses. Our acoustic word embeddings improve over previous approaches for the task of word discrimination. We also present results on other tasks that are enabled by the multi-view approach, including cross-view word discrimination and word similarity.

Hao Tang, Weiran Wang, Kevin Gimpel et al.

Recent work on discriminative segmental models has shown that they can achieve competitive speech recognition performance, using features based on deep neural frame classifiers. However, segmental models can be more challenging to train than standard frame-based approaches. While some segmental models have been successfully trained end to end, there is a lack of understanding of their training under different settings and with different losses. We investigate a model class based on recent successful approaches, consisting of a linear model that combines segmental features based on an LSTM frame classifier. Similarly to hybrid HMM-neural network models, segmental models of this class can be trained in two stages (frame classifier training followed by linear segmental model weight training), end to end (joint training of both frame classifier and linear weights), or with end-to-end fine-tuning after two-stage training. We study segmental models trained end to end with hinge loss, log loss, latent hinge loss, and marginal log loss. We consider several losses for the case where training alignments are available as well as where they are not. We find that in general, marginal log loss provides the most consistent strong performance without requiring ground-truth alignments. We also find that training with dropout is very important in obtaining good performance with end-to-end training. Finally, the best results are typically obtained by a combination of two-stage training and fine-tuning.

Weiran Wang, Xinchen Yan, Honglak Lee et al.

We present deep variational canonical correlation analysis (VCCA), a deep multi-view learning model that extends the latent variable model interpretation of linear CCA to nonlinear observation models parameterized by deep neural networks. We derive variational lower bounds of the data likelihood by parameterizing the posterior probability of the latent variables from the view that is available at test time. We also propose a variant of VCCA called VCCA-private that can, in addition to the "common variables" underlying both views, extract the "private variables" within each view, and disentangles the shared and private information for multi-view data without hard supervision. Experimental results on real-world datasets show that our methods are competitive across domains.

Taehwan Kim, Jonathan Keane, Weiran Wang et al.

We study the problem of recognizing video sequences of fingerspelled letters in American Sign Language (ASL). Fingerspelling comprises a significant but relatively understudied part of ASL. Recognizing fingerspelling is challenging for a number of reasons: It involves quick, small motions that are often highly coarticulated; it exhibits significant variation between signers; and there has been a dearth of continuous fingerspelling data collected. In this work we collect and annotate a new data set of continuous fingerspelling videos, compare several types of recognizers, and explore the problem of signer variation. Our best-performing models are segmental (semi-Markov) conditional random fields using deep neural network-based features. In the signer-dependent setting, our recognizers achieve up to about 92% letter accuracy. The multi-signer setting is much more challenging, but with neural network adaptation we achieve up to 83% letter accuracies in this setting.