Searcharxiv⌕ Search

arXiv subjects

Liang Lu

Publications and source records attributed to Liang Lu.

54 records · Page 3Linked to original sources
arXiv2019

Self-Teaching Networks

We propose self-teaching networks to improve the generalization capacity of deep neural networks. The idea is to generate soft supervision labels using the output layer for training the lower layers of the network. During the network training, we seek an auxiliary loss that drives the lower layer to mimic the behavior of the output layer. The connection between the two network layers through the auxiliary loss can help the gradient flow, which works similar to the residual networks. Furthermore, the auxiliary loss also works as a regularizer, which improves the generalization capacity of the network. We evaluated the self-teaching network with deep recurrent neural networks on speech recognition tasks, where we trained the acoustic model using 30 thousand hours of data. We tested the acoustic model using data collected from 4 scenarios. We show that the self-teaching network can achieve consistent improvements and outperform existing methods such as label smoothing and confidence penalization.

eess.AS↗
arXiv2018

A Study of All-Convolutional Encoders for Connectionist Temporal Classification

Connectionist temporal classification (CTC) is a popular sequence prediction approach for automatic speech recognition that is typically used with models based on recurrent neural networks (RNNs). We explore whether deep convolutional neural networks (CNNs) can be used effectively instead of RNNs as the "encoder" in CTC. CNNs lack an explicit representation of the entire sequence, but have the advantage that they are much faster to train. We present an exploration of CNNs as encoders for CTC models, in the context of character-based (lexicon-free) automatic speech recognition. In particular, we explore a range of one-dimensional convolutional layers, which are particularly efficient. We compare the performance of our CNN-based models against typical RNNbased models in terms of training time, decoding time, model size and word error rate (WER) on the Switchboard Eval2000 corpus. We find that our CNN-based models are close in performance to LSTMs, while not matching them, and are much faster to train and decode.

cs.CL↗
arXiv2017

Multiplicative LSTM for sequence modelling

We introduce multiplicative LSTM (mLSTM), a recurrent neural network architecture for sequence modelling that combines the long short-term memory (LSTM) and multiplicative recurrent neural network architectures. mLSTM is characterised by its ability to have different recurrent transition functions for each possible input, which we argue makes it more expressive for autoregressive density estimation. We demonstrate empirically that mLSTM outperforms standard LSTM and its deep variants for a range of character level language modelling tasks. In this version of the paper, we regularise mLSTM to achieve 1.27 bits/char on text8 and 1.24 bits/char on Hutter Prize. We also apply a purely byte-level mLSTM on the WikiText-2 dataset to achieve a character level entropy of 1.26 bits/char, corresponding to a word level perplexity of 88.8, which is comparable to word level LSTMs regularised in similar ways on the same task.

cs.NE↗
arXiv2017

End-to-End Neural Segmental Models for Speech Recognition

Segmental models are an alternative to frame-based models for sequence prediction, where hypothesized path weights are based on entire segment scores rather than a single frame at a time. Neural segmental models are segmental models that use neural network-based weight functions. Neural segmental models have achieved competitive results for speech recognition, and their end-to-end training has been explored in several studies. In this work, we review neural segmental models, which can be viewed as consisting of a neural network-based acoustic encoder and a finite-state transducer decoder. We study end-to-end segmental models with different weight functions, including ones based on frame-level neural classifiers and on segmental recurrent neural networks. We study how reducing the search space size impacts performance under different weight functions. We also compare several loss functions for end-to-end training. Finally, we explore training approaches, including multi-stage vs. end-to-end training and multitask training that combines segmental and frame-level losses.

cs.CL↗
arXiv2017

Toward Computation and Memory Efficient Neural Network Acoustic Models with Binary Weights and Activations

Neural network acoustic models have significantly advanced state of the art speech recognition over the past few years. However, they are usually computationally expensive due to the large number of matrix-vector multiplications and nonlinearity operations. Neural network models also require significant amounts of memory for inference because of the large model size. For these two reasons, it is challenging to deploy neural network based speech recognizers on resource-constrained platforms such as embedded devices. This paper investigates the use of binary weights and activations for computation and memory efficient neural network acoustic models. Compared to real-valued weight matrices, binary weights require much fewer bits for storage, thereby cutting down the memory footprint. Furthermore, with binary weights or activations, the matrix-vector multiplications are turned into addition and subtraction operations, which are computationally much faster and more energy efficient for hardware platforms. In this paper, we study the applications of binary weights and activations for neural network acoustic modeling, reporting encouraging results on the WSJ and AMI corpora.

cs.CL↗
arXiv2017

Small-footprint Deep Neural Networks with Highway Connections for Speech Recognition

For speech recognition, deep neural networks (DNNs) have significantly improved the recognition accuracy in most of benchmark datasets and application domains. However, compared to the conventional Gaussian mixture models, DNN-based acoustic models usually have much larger number of model parameters, making it challenging for their applications in resource constrained platforms, e.g., mobile devices. In this paper, we study the application of the recently proposed highway network to train small-footprint DNNs, which are {\it thinner} and {\it deeper}, and have significantly smaller number of model parameters compared to conventional DNNs. We investigated this approach on the AMI meeting speech transcription corpus which has around 70 hours of audio data. The highway neural networks constantly outperformed their plain DNN counterparts, and the number of model parameters can be reduced significantly without sacrificing the recognition accuracy.

cs.CL↗
arXiv2017

Multitask Learning with CTC and Segmental CRF for Speech Recognition

Segmental conditional random fields (SCRFs) and connectionist temporal classification (CTC) are two sequence labeling methods used for end-to-end training of speech recognition models. Both models define a transcription probability by marginalizing decisions about latent segmentation alternatives to derive a sequence probability: the former uses a globally normalized joint model of segment labels and durations, and the latter classifies each frame as either an output symbol or a "continuation" of the previous label. In this paper, we train a recognition model by optimizing an interpolation between the SCRF and CTC losses, where the same recurrent neural network (RNN) encoder is used for feature extraction for both outputs. We find that this multitask objective improves recognition accuracy when decoding with either the SCRF or CTC models. Additionally, we show that CTC can also be used to pretrain the RNN encoder, which improves the convergence rate when learning the joint model.

cs.CL↗
arXiv2017

Adaptive DCTNet for Audio Signal Classification

In this paper, we investigate DCTNet for audio signal classification. Its output feature is related to Cohen's class of time-frequency distributions. We introduce the use of adaptive DCTNet (A-DCTNet) for audio signals feature extraction. The A-DCTNet applies the idea of constant-Q transform, with its center frequencies of filterbanks geometrically spaced. The A-DCTNet is adaptive to different acoustic scales, and it can better capture low frequency acoustic information that is sensitive to human audio perception than features such as Mel-frequency spectral coefficients (MFSC). We use features extracted by the A-DCTNet as input for classifiers. Experimental results show that the A-DCTNet and Recurrent Neural Networks (RNN) achieve state-of-the-art performance in bird song classification rate, and improve artist identification accuracy in music data. They demonstrate A-DCTNet's applicability to signal processing problems.

cs.SD↗
arXiv2017

Small-footprint Highway Deep Neural Networks for Speech Recognition

State-of-the-art speech recognition systems typically employ neural network acoustic models. However, compared to Gaussian mixture models, deep neural network (DNN) based acoustic models often have many more model parameters, making it challenging for them to be deployed on resource-constrained platforms, such as mobile devices. In this paper, we study the application of the recently proposed highway deep neural network (HDNN) for training small-footprint acoustic models. HDNNs are a depth-gated feedforward neural network, which include two types of gate functions to facilitate the information flow through different layers. Our study demonstrates that HDNNs are more compact than regular DNNs for acoustic modeling, i.e., they can achieve comparable recognition accuracy with many fewer model parameters. Furthermore, HDNNs are more controllable than DNNs: the gate functions of an HDNN can control the behavior of the whole network using a very small number of model parameters. Finally, we show that HDNNs are more adaptable than DNNs. For example, simply updating the gate functions using adaptation data can result in considerable gains in accuracy. We demonstrate these aspects by experiments using the publicly available AMI corpus, which has around 80 hours of training data.

cs.CL↗
arXiv2017

Multitask Learning with Low-Level Auxiliary Tasks for Encoder-Decoder Based Speech Recognition

End-to-end training of deep learning-based models allows for implicit learning of intermediate representations based on the final task loss. However, the end-to-end approach ignores the useful domain knowledge encoded in explicit intermediate-level supervision. We hypothesize that using intermediate representations as auxiliary supervision at lower levels of deep networks may be a good way of combining the advantages of end-to-end training and more traditional pipeline approaches. We present experiments on conversational speech recognition where we use lower-level tasks, such as phoneme recognition, in a multitask training approach with an encoder-decoder model for direct character transcription. We compare multiple types of lower-level tasks and analyze the effects of the auxiliary tasks. Our results on the Switchboard corpus show that this approach improves recognition accuracy over a standard encoder-decoder model on the Eval2000 test set.

cs.CL↗
arXiv2017

Sequence Training and Adaptation of Highway Deep Neural Networks

Highway deep neural network (HDNN) is a type of depth-gated feedforward neural network, which has shown to be easier to train with more hidden layers and also generalise better compared to conventional plain deep neural networks (DNNs). Previously, we investigated a structured HDNN architecture for speech recognition, in which the two gate functions were tied across all the hidden layers, and we were able to train a much smaller model without sacrificing the recognition accuracy. In this paper, we carry on the study of this architecture with sequence-discriminative training criterion and speaker adaptation techniques on the AMI meeting speech recognition corpus. We show that these two techniques improve speech recognition accuracy on top of the model trained with the cross entropy criterion. Furthermore, we demonstrate that the two gate functions that are tied across all the hidden layers are able to control the information flow over the whole network, and we can achieve considerable improvements by only updating these gate functions in both sequence training and adaptation experiments.

cs.CL↗
arXiv2016

Knowledge Distillation for Small-footprint Highway Networks

Deep learning has significantly advanced state-of-the-art of speech recognition in the past few years. However, compared to conventional Gaussian mixture acoustic models, neural network models are usually much larger, and are therefore not very deployable in embedded devices. Previously, we investigated a compact highway deep neural network (HDNN) for acoustic modelling, which is a type of depth-gated feedforward neural network. We have shown that HDNN-based acoustic models can achieve comparable recognition accuracy with much smaller number of model parameters compared to plain deep neural network (DNN) acoustic models. In this paper, we push the boundary further by leveraging on the knowledge distillation technique that is also known as {\it teacher-student} training, i.e., we train the compact HDNN model with the supervision of a high accuracy cumbersome model. Furthermore, we also investigate sequence training and adaptation in the context of teacher-student training. Our experiments were performed on the AMI meeting speech recognition corpus. With this technique, we significantly improved the recognition accuracy of the HDNN acoustic model with less than 0.8 million parameters, and narrowed the gap between this model and the plain DNN with 30 million parameters.

cs.CL↗
arXiv2016

Segmental Recurrent Neural Networks for End-to-end Speech Recognition

We study the segmental recurrent neural network for end-to-end acoustic modelling. This model connects the segmental conditional random field (CRF) with a recurrent neural network (RNN) used for feature extraction. Compared to most previous CRF-based acoustic models, it does not rely on an external system to provide features or segmentation boundaries. Instead, this model marginalises out all the possible segmentations, and features are extracted from the RNN trained together with the segmental CRF. In essence, this model is self-contained and can be trained end-to-end. In this paper, we discuss practical training and decoding issues as well as the method to speed up the training in the context of speech recognition. We performed experiments on the TIMIT dataset. We achieved 17.3 phone error rate (PER) from the first-pass decoding --- the best reported result using CRFs, despite the fact that we only used a zeroth-order CRF and without using any language model.

cs.CL↗
arXiv2016

Top-down Tree Long Short-Term Memory Networks

Long Short-Term Memory (LSTM) networks, a type of recurrent neural network with a more complex computational unit, have been successfully applied to a variety of sequence modeling tasks. In this paper we develop Tree Long Short-Term Memory (TreeLSTM), a neural network model based on LSTM, which is designed to predict a tree rather than a linear sequence. TreeLSTM defines the probability of a sentence by estimating the generation probability of its dependency tree. At each time step, a node is generated based on the representation of the generated sub-tree. We further enhance the modeling power of TreeLSTM by explicitly representing the correlations between left and right dependents. Application of our model to the MSR sentence completion challenge achieves results beyond the current state of the art. We also report results on dependency parsing reranking achieving competitive performance.

cs.CL↗
arXiv2015

Successful High Power Acceleration of the HSC Type Injector for Cancer Therapy in IMP

A Hybrid single cavity (HSC) linac, which is formed by combining a radio frequency quadrupole (RFQ) structure and a drift tube (DT) structure into one interdigital-H (IH) cavity, is fabricated and assembled as a proof of principle (PoP) type injector for cancer therapy synchrotron according to the series researches. The injection method of the HSC linac adopt a direct plasma injection scheme (DPIS), which is considered to be the only method for accelerating an intense current heavy ion beam produced by a laser ion source (LIS). The input beam current of the HSC was designed to be 20 milliampere (mA) C6+ ions. According to numerical simulations, the HSC linac could accelerate a 6-mA C6+ beam which meets requirement of the needed particle number for cancer therapy use (108~9 ions per pulse). The injection system adopted HSC injector with the method of DPIS can make the existing multi-turn injection system and the stripping system unnecessary, and also can make the beam pipe of the existing synchrotron magnets downsize which could reduce the whole cost of synchrotron (estimated millions dollars). Details of the measurements and evaluations of the assembled HSC linac, and the results of the high power test using a LIS with DPIS are reported in this paper.

physics.acc-ph↗
arXiv2014

Tied Probabilistic Linear Discriminant Analysis for Speech Recognition

Acoustic models using probabilistic linear discriminant analysis (PLDA) capture the correlations within feature vectors using subspaces which do not vastly expand the model. This allows high dimensional and correlated feature spaces to be used, without requiring the estimation of multiple high dimension covariance matrices. In this letter we extend the recently presented PLDA mixture model for speech recognition through a tied PLDA approach, which is better able to control the model size to avoid overfitting. We carried out experiments using the Switchboard corpus, with both mel frequency cepstral coefficient features and bottleneck feature derived from a deep neural network. Reductions in word error rate were obtained by using tied PLDA, compared with the PLDA mixture model, subspace Gaussian mixture models, and deep neural networks.

cs.CL↗
arXiv2005

Significance Calculation and a New Analysis Method in Searching for New Physics at the LHC

The LHC experiments have great potential in discovering many possible new particles up to the TeV scale. The significance calculation of an observation of a physics signal with known location and shape is no longer valid when either the location or the shape of the signal is unknown. We find the current LHC significance calculation of new physics is over-estimated and strongly depends on the specifics of the method and the situation it applies to. We describe general procedures for significance calculation and comparing different search schemes. A new method uses maximum likelihood fits with floating parameters and scans the parameter space for the best fit to the entire sample. We find that the new method is significantly more sensitive than current method and is insensitive to the exact location of the new physics signal we search.

physics.ins-det↗