Research papers and code for "Dong Yu":
Recent breakthroughs in Neural Architectural Search (NAS) have achieved state-of-the-art performance in many tasks such as image classification and language understanding. However, most existing works only optimize for model accuracy and largely ignore other important factors imposed by the underlying hardware and devices, such as latency and energy, when making inference. In this paper, we first introduce the problem of NAS and provide a survey on recent works. Then we deep dive into two recent advancements on extending NAS into multiple-objective frameworks: MONAS and DPP-Net. Both MONAS and DPP-Net are capable of optimizing accuracy and other objectives imposed by devices, searching for neural architectures that can be best deployed on a wide spectrum of devices: from embedded systems and mobile devices to workstations. Experimental results are poised to show that architectures found by MONAS and DPP-Net achieves Pareto optimality w.r.t the given objectives for various devices.

* ICCAD'18 Invited Paper
Click to Read Paper and Get Code
Support vector clustering (SVC) is a versatile clustering technique that is able to identify clusters of arbitrary shapes by exploiting the kernel trick. However, one hurdle that restricts the application of SVC lies in its sensitivity to the kernel parameter and the trade-off parameter. Although many extensions of SVC have been developed, to the best of our knowledge, there is still no algorithm that is able to effectively estimate the two crucial parameters in SVC without supervision. In this paper, we propose a novel support vector clustering approach termed ensemble-driven support vector clustering (EDSVC), which for the first time tackles the automatic parameter estimation problem for SVC based on ensemble learning, and is capable of producing robust clustering results in a purely unsupervised manner. Experimental results on multiple real-world datasets demonstrate the effectiveness of our approach.

* To appear in ICPR 2016
Click to Read Paper and Get Code
In this work, three lattice-free (LF) discriminative training criteria for purely sequence-trained neural network acoustic models are compared on LVCSR tasks, namely maximum mutual information (MMI), boosted maximum mutual information (bMMI) and state-level minimum Bayes risk (sMBR). We demonstrate that, analogous to LF-MMI, a neural network acoustic model can also be trained from scratch using LF-bMMI or LF-sMBR criteria respectively without the need of cross-entropy pre-training. Furthermore, experimental results on Switchboard-300hrs and Switchboard+Fisher-2100hrs datasets show that models trained with LF-bMMI consistently outperform those trained with plain LF-MMI and achieve a relative word error rate (WER) reduction of 5% over competitive temporal convolution projected LSTM (TDNN-LSTMP) LF-MMI baselines.

* under review ICASSP2019
Click to Read Paper and Get Code
In this paper, we summarize recent progresses made in deep learning based acoustic models and the motivation and insights behind the surveyed techniques. We first discuss acoustic models that can effectively exploit variable-length contextual information, such as recurrent neural networks (RNNs), convolutional neural networks (CNNs), and their various combination with other models. We then describe acoustic models that are optimized end-to-end with emphasis on feature representations learned jointly with rest of the system, the connectionist temporal classification (CTC) criterion, and the attention-based sequence-to-sequence model. We further illustrate robustness issues in speech recognition systems, and discuss acoustic model adaptation, speech enhancement and separation, and robust training strategies. We also cover modeling techniques that lead to more efficient decoding and discuss possible future directions in acoustic model research.

* This is an updated version with latest literature until ICASSP2018 of the paper: Dong Yu and Jinyu Li, "Recent Progresses in Deep Learning based Acoustic Models," vol.4, no.3, IEEE/CAA Journal of Automatica Sinica, 2017
Click to Read Paper and Get Code
In many automatic speech recognition (ASR) tasks, an ideal model has to be applicable over multiple domains. In this paper, we propose to teach an all-rounder with experts in different domains. Concretely, we build a multi-domain acoustic model by applying the teacher-student training framework. First, for each domain, a teacher model (domain-dependent model) is trained by fine-tuning a multi-condition model with domain-specific subset. Then all these teacher models are used to teach one single student model simultaneously. We perform experiments on two predefined domain setups. One is domains with different speaking styles, the other is nearfield, far-field and far-field with noise. Moreover, two types of models are examined: deep feedforward sequential memory network (DFSMN) and long short term memory (LSTM). Experimental results show that the model trained with this framework outperforms not only multi-condition model but also domain-dependent model. Specially, our training method provides up to 10.4% relative character error rate improvement over baseline model (multi-condition model).

* 5 pages and 2 figures; accepted by 2019 IEEE International Conference on Acoustics, Speech and Signal Processing
Click to Read Paper and Get Code
In image restoration tasks, like denoising and super resolution, continual modulation of restoration levels is of great importance for real-world applications, but has failed most of existing deep learning based image restoration methods. Learning from discrete and fixed restoration levels, deep models cannot be easily generalized to data of continuous and unseen levels. This topic is rarely touched in literature, due to the difficulty of modulating well-trained models with certain hyper-parameters. We make a step forward by proposing a unified CNN framework that consists of few additional parameters than a single-level model yet could handle arbitrary restoration levels between a start and an end level. The additional module, namely AdaFM layer, performs channel-wise feature modification, and can adapt a model to another restoration level with high accuracy. By simply tweaking an interpolation coefficient, the intermediate model - AdaFM-Net could generate smooth and continuous restoration effects without artifacts. Extensive experiments on three image restoration tasks demonstrate the effectiveness of both model training and modulation testing. Besides, we carefully investigate the properties of AdaFM layers, providing a detailed guidance on the usage of the proposed method.

* Accepted by CVPR 2019 (oral)
Click to Read Paper and Get Code
Word embeddings are traditionally trained on a large corpus in an unsupervised setting, with no specific design for incorporating domain knowledge. This can lead to unsatisfactory performances when training data originate from heterogeneous domains. In this paper, we propose two novel mechanisms for domain-aware word embedding training, namely domain indicator and domain attention, which integrate domain-specific knowledge into the widely used SG and CBOW models, respectively. The two methods are based on a joint learning paradigm and ensure that words in a target domain are intensively focused when trained on a source domain corpus. Qualitative and quantitative evaluation confirm the validity and effectiveness of our models. Compared to baseline methods, our method is particularly effective in near-cold-start scenarios.

Click to Read Paper and Get Code
In this paper, we propose an improvement of Adversarial Transformation Networks(ATN) to generate adversarial examples, which can fool white-box models and black-box models with a state of the art performance and won the 2rd place in the non-target task in CAAD 2018.

Click to Read Paper and Get Code
Sequence-to-sequence (seq2seq) neural models have been actively investigated for abstractive summarization. Nevertheless, existing neural abstractive systems frequently generate factually incorrect summaries and are vulnerable to adversarial information, suggesting a crucial lack of semantic understanding. In this paper, we propose a novel semantic-aware neural abstractive summarization model that learns to generate high quality summaries through semantic interpretation over salient content. A novel evaluation scheme with adversarial samples is introduced to measure how well a model identifies off-topic information, where our model yields significantly better performance than the popular pointer-generator summarizer. Human evaluation also confirms that our system summaries are uniformly more informative and faithful as well as less redundant than the seq2seq model.

Click to Read Paper and Get Code
With the excellent accuracy and feasibility, the Neural Networks have been widely applied into the novel intelligent applications and systems. However, with the appearance of the Adversarial Attack, the NN based system performance becomes extremely vulnerable:the image classification results can be arbitrarily misled by the adversarial examples, which are crafted images with human unperceivable pixel-level perturbation. As this raised a significant system security issue, we implemented a series of investigations on the adversarial attack in this work: We first identify an image's pixel vulnerability to the adversarial attack based on the adversarial saliency analysis. By comparing the analyzed saliency map and the adversarial perturbation distribution, we proposed a new evaluation scheme to comprehensively assess the adversarial attack precision and efficiency. Then, with a novel adversarial saliency prediction method, a fast adversarial example generation framework, namely "ASP", is proposed with significant attack efficiency improvement and dramatic computation cost reduction. Compared to the previous methods, experiments show that ASP has at most 12 times speed-up for adversarial example generation, 2 times lower perturbation rate, and high attack success rate of 87% on both MNIST and Cifar10. ASP can be also well utilized to support the data-hungry NN adversarial training. By reducing the attack success rate as much as 90%, ASP can quickly and effectively enhance the defense capability of NN based system to the adversarial attacks.

Click to Read Paper and Get Code
Although great progresses have been made in automatic speech recognition (ASR), significant performance degradation is still observed when recognizing multi-talker mixed speech. In this paper, we propose and evaluate several architectures to address this problem under the assumption that only a single channel of mixed signal is available. Our technique extends permutation invariant training (PIT) by introducing the front-end feature separation module with the minimum mean square error (MSE) criterion and the back-end recognition module with the minimum cross entropy (CE) criterion. More specifically, during training we compute the average MSE or CE over the whole utterance for each possible utterance-level output-target assignment, pick the one with the minimum MSE or CE, and optimize for that assignment. This strategy elegantly solves the label permutation problem observed in the deep learning based multi-talker mixed speech separation and recognition systems. The proposed architectures are evaluated and compared on an artificially mixed AMI dataset with both two- and three-talker mixed speech. The experimental results indicate that our proposed architectures can cut the word error rate (WER) by 45.0% and 25.0% relatively against the state-of-the-art single-talker speech recognition system across all speakers when their energies are comparable, for two- and three-talker mixed speech, respectively. To our knowledge, this is the first work on the multi-talker mixed speech recognition on the challenging speaker-independent spontaneous large vocabulary continuous speech task.

* 11 pages, 6 figures, Submitted to IEEE/ACM Transactions on Audio, Speech and Language Processing. arXiv admin note: text overlap with arXiv:1704.01985
Click to Read Paper and Get Code
In this paper, we propose a novel technique for direct recognition of multiple speech streams given the single channel of mixed speech, without first separating them. Our technique is based on permutation invariant training (PIT) for automatic speech recognition (ASR). In PIT-ASR, we compute the average cross entropy (CE) over all frames in the whole utterance for each possible output-target assignment, pick the one with the minimum CE, and optimize for that assignment. PIT-ASR forces all the frames of the same speaker to be aligned with the same output layer. This strategy elegantly solves the label permutation problem and speaker tracing problem in one shot. Our experiments on artificially mixed AMI data showed that the proposed approach is very promising.

* 5 pages, 6 figures, InterSpeech2017
Click to Read Paper and Get Code
Resolving pronoun coreference requires knowledge support, especially for particular domains (e.g., medicine). In this paper, we explore how to leverage different types of knowledge to better resolve pronoun coreference with a neural model. To ensure the generalization ability of our model, we directly incorporate knowledge in the format of triplets, which is the most common format of modern knowledge graphs, instead of encoding it with features or rules as that in conventional approaches. Moreover, since not all knowledge is helpful in certain contexts, to selectively use them, we propose a knowledge attention module, which learns to select and use informative knowledge based on contexts, to enhance our model. Experimental results on two datasets from different domains prove the validity and effectiveness of our model, where it outperforms state-of-the-art baselines by a large margin. Moreover, since our model learns to use external knowledge rather than only fitting the training data, it also demonstrates superior performance to baselines in the cross-domain setting.

* Accepted by ACL 2019
Click to Read Paper and Get Code
The cloud-based speech recognition/API provides developers or enterprises an easy way to create speech-enabled features in their applications. However, sending audios about personal or company internal information to the cloud, raises concerns about the privacy and security issues. The recognition results generated in cloud may also reveal some sensitive information. This paper proposes a deep polynomial network (DPN) that can be applied to the encrypted speech as an acoustic model. It allows clients to send their data in an encrypted form to the cloud to ensure that their data remains confidential, at mean while the DPN can still make frame-level predictions over the encrypted speech and return them in encrypted form. One good property of the DPN is that it can be trained on unencrypted speech features in the traditional way. To keep the cloud away from the raw audio and recognition results, a cloud-local joint decoding framework is also proposed. We demonstrate the effectiveness of model and framework on the Switchboard and Cortana voice assistant tasks with small performance degradation and latency increased comparing with the traditional cloud-based DNNs.

* ICASSP 2019, slides@ https://www.researchgate.net/publication/333005422_Encrypted_Speech_Recognition_using_deep_polynomial_networks
Click to Read Paper and Get Code
With an ultimate goal of narrowing the gap between human and machine readers in text comprehension, we present the first collection of Challenging Chinese machine reading Comprehension datasets (C^3) collected from language and professional certification exams, which contains 13,924 documents and their associated 23,990 multiple-choice questions. Most of the questions in C^3 cannot be answered merely by surface-form matching against the given text. As a pilot study, we closely analyze the prior knowledge (i.e., linguistic, domain-specific, and general world knowledge) needed in these real-world reading comprehension tasks. We further explore how to leverage linguistic knowledge including a lexicon of idioms and proverbs, graphs of general world knowledge (e.g., ConceptNet), and domain-specific knowledge such as textbooks to aid machine readers, through fine-tuning a pre-trained language model. Experimental results demonstrate that linguistic and general world knowledge may help improve the performance of the baseline reader in both general and domain-specific tasks. C^3 will be available at http://dataset.org/c3/.

Click to Read Paper and Get Code
We aim to study the multi-scale receptive fields of a single convolutional neural network to detect faces of varied scales. This paper presents our Multi-Scale Receptive Field Face Detector (MSFD), which has superior performance on detecting faces at different scales and enjoys real-time inference speed. MSFD agglomerates context and texture by hierarchical structure. More additional information and rich receptive field bring significant improvement but generate marginal time consumption. We simultaneously propose an anchor assignment strategy which can cover faces with a wide range of scales to improve the recall rate of small faces and rotated faces. To reduce the false positive rate, we train our detector with focal loss which keeps the easy samples from overwhelming. As a result, MSFD reaches superior results on the FDDB, Pascal-Faces and WIDER FACE datasets, and can run at 31 FPS on GPU for VGA-resolution images.

* Accepted by ICPR 2018
Click to Read Paper and Get Code
Natural Language Inference (NLI) is a fundamental and challenging task in Natural Language Processing (NLP). Most existing methods only apply one-pass inference process on a mixed matching feature, which is a concatenation of different matching features between a premise and a hypothesis. In this paper, we propose a new model called Multi-turn Inference Matching Network (MIMN) to perform multi-turn inference on different matching features. In each turn, the model focuses on one particular matching feature instead of the mixed matching feature. To enhance the interaction between different matching features, a memory component is employed to store the history inference information. The inference of each turn is performed on the current matching feature and the memory. We conduct experiments on three different NLI datasets. The experimental results show that our model outperforms or achieves the state-of-the-art performance on all the three datasets.

Click to Read Paper and Get Code
Reading strategies have been shown to improve comprehension levels, especially for readers lacking adequate prior knowledge. Just as the process of knowledge accumulation is time-consuming for human readers, it is resource-demanding to impart rich general domain knowledge into a language model via pre-training (Radford et al., 2018; Devlin et al., 2018). Inspired by reading strategies identified in cognitive science, and given limited computational resources - just a pre-trained model and a fixed number of training instances - we therefore propose three simple domain-independent strategies aimed to improve non-extractive machine reading comprehension (MRC): (i) BACK AND FORTH READING that considers both the original and reverse order of an input sequence, (ii) HIGHLIGHTING, which adds a trainable embedding to the text embedding of tokens that are relevant to the question and candidate answers, and (iii) SELF-ASSESSMENT that generates practice questions and candidate answers directly from the text in an unsupervised manner. By fine-tuning a pre-trained language model (Radford et al., 2018) with our proposed strategies on the largest existing general domain multiple-choice MRC dataset RACE, we obtain a 5.8% absolute increase in accuracy over the previous best result achieved by the same pre-trained model fine-tuned on RACE without the use of strategies. We further fine-tune the resulting model on a target task, leading to new state-of-the-art results on six representative non-extractive MRC datasets from different domains (i.e., ARC, OpenBookQA, MCTest, MultiRC, SemEval-2018, and ROCStories). These results indicate the effectiveness of the proposed strategies and the versatility and general applicability of our fine-tuned models that incorporate these strategies.

Click to Read Paper and Get Code