arXiv subjects

Hao Zhou

Publications and source records attributed to Hao Zhou.

At least 415 records · Page 23Linked to original sources

arXiv2021

Physiological-Physical Feature Fusion for Automatic Voice Spoofing Detection

Speaker verification systems have been used in many production scenarios in recent years. Unfortunately, they are still highly prone to different kinds of spoofing attacks such as voice conversion and speech synthesis, etc. In this paper, we propose a new method base on physiological-physical feature fusion to deal with voice spoofing attacks. This method involves feature extraction, a densely connected convolutional neural network with squeeze and excitation block (SE-DenseNet), multi-scale residual neural network with squeeze and excitation block (SE-Res2Net) and feature fusion strategies. We first pre-trained a convolutional neural network using the speaker's voice and face in the video as surveillance signals. It can extract physiological features from speech. Then we use SE-DenseNet and SE-Res2Net to extract physical features. Such a densely connection pattern has high parameter efficiency and squeeze and excitation block can enhance the transmission of the feature. Finally, we integrate the two features into the SE-Densenet to identify the spoofing attacks. Experimental results on the ASVspoof 2019 data set show that our model is effective for voice spoofing detection. In the logical access scenario, our model improves the tandem decision cost function (t-DCF) and equal error rate (EER) scores by 4% and 7%, respectively, compared with other methods. In the physical access scenario, our model improved t-DCF and EER scores by 8% and 10%, respectively.

eess.AS↗

arXiv2021

Impact of Acceleration/deceleration Limits on the String Stability of Adaptive Cruise Control

This paper demonstrates that the acceleration/deceleration limits in ACC systems can make a string stable ACC amplify the speed perturbation in natural driving. It is shown that the constrained acceleration/deceleration of the following ACCs are likely to cause speed overshoot to compensate for an extra large/small spacing. Additionally, we find that the constrained deceleration limits can also jeopardize safety, as the limited braking power produces extra small spacing or even crashes. The findings are validated through experiments on real cars. The paper suggests that the ACC parameter space should be extended to include the acceleration/deceleration limits considering their significant role exposed here. Through numerical simulations of ACC platoons, we show i) a marginal string stable ACC is preferable due to the smaller total queue length and the shorter duration in congestion; ii) congestion waves in a mixed ACC platoon largely depend on the sequence of vehicles provided different acceleration/deceleration limits, and iii) the safety hazard caused by the constrained deceleration limits is more significant in mixed ACC platoons when string unstable ACCs exist.

eess.SY↗

arXiv2021

EVA: An Open-Domain Chinese Dialogue System with Large-Scale Generative Pre-Training

Although pre-trained language models have remarkably enhanced the generation ability of dialogue systems, open-domain Chinese dialogue systems are still limited by the dialogue data and the model size compared with English ones. In this paper, we propose EVA, a Chinese dialogue system that contains the largest Chinese pre-trained dialogue model with 2.8B parameters. To build this model, we collect the largest Chinese dialogue dataset named WDC-Dialogue from various public social media. This dataset contains 1.4B context-response pairs and is used as the pre-training corpus of EVA. Extensive experiments on automatic and human evaluation show that EVA outperforms other Chinese pre-trained dialogue models especially in the multi-turn interaction of human-bot conversations.

cs.CL↗

arXiv2021

Follow Your Path: a Progressive Method for Knowledge Distillation

Deep neural networks often have a huge number of parameters, which posts challenges in deployment in application scenarios with limited memory and computation capacity. Knowledge distillation is one approach to derive compact models from bigger ones. However, it has been observed that a converged heavy teacher model is strongly constrained for learning a compact student network and could make the optimization subject to poor local optima. In this paper, we propose ProKT, a new model-agnostic method by projecting the supervision signals of a teacher model into the student's parameter space. Such projection is implemented by decomposing the training objective into local intermediate targets with an approximate mirror descent technique. The proposed method could be less sensitive with the quirks during optimization which could result in a better local optimum. Experiments on both image and text datasets show that our proposed ProKT consistently achieves superior performance compared to other existing knowledge distillation methods.

cs.LG↗

arXiv2021

Review of Learning-based Longitudinal Motion Planning for Autonomous Vehicles: Research Gaps between Self-driving and Traffic Congestion

Self-driving technology companies and the research community are accelerating their pace to use machine learning longitudinal motion planning (mMP) for autonomous vehicles (AVs). This paper reviews the current state of the art in mMP, with an exclusive focus on its impact on traffic congestion. We identify the availability of congestion scenarios in current datasets, and summarize the required features for training mMP. For learning methods, we survey the major methods in both imitation learning and non-imitation learning. We also highlight the emerging technologies adopted by some leading AV companies, e.g. Tesla, Waymo, and Comma.ai. We find that: i) the AV industry has been mostly focusing on the long tail problem related to safety and overlooked the impact on traffic congestion, ii) the current public self-driving datasets have not included enough congestion scenarios, and mostly lack the necessary input features/output labels to train mMP, and iii) albeit reinforcement learning (RL) approach can integrate congestion mitigation into the learning goal, the major mMP method adopted by industry is still behavior cloning (BC), whose capability to learn a congestion-mitigating mMP remains to be seen. Based on the review, the study identifies the research gaps in current mMP development. Some suggestions towards congestion mitigation for future mMP studies are proposed: i) enrich data collection to facilitate the congestion learning, ii) incorporate non-imitation learning methods to combine traffic efficiency into a safety-oriented technical route, and iii) integrate domain knowledge from the traditional car following (CF) theory to improve the string stability of mMP.

eess.SP↗

arXiv2021

Significance of Low-level Controller for String Stability under Adaptive Cruise Control

Current commercial adaptive cruise control (ACC) systems consist of an upper-level planner controller that decides the optimal trajectory that should be followed, and a low-level controller in charge of sending the gas/brake signals to the mechanical system to actually move the vehicle. We find that the low-level controller has a significant impact on the string stability (SS) even if the planner is string stable: (i) a slow controller deteriorates the SS, (ii) slow controllers are common as they arise from insufficient control gains, from a "weak" gas/brake system or both, and (iii) the integral term in a slow controller causes undesired overshooting which affects the SS. Accordingly, we suggest tuning up the proportional/feedforward gain and ensuring the gas/brake is not "weak". The study results are validated both numerically and empirically with data from commercial cars.

eess.SY↗

arXiv2021

Financial Network Games

We study financial systems from a game-theoretic standpoint. A financial system is represented by a network, where nodes correspond to firms, and directed labeled edges correspond to debt contracts between them. The existence of cycles in the network indicates that a payment of a firm to one of its lenders might result to some incoming payment. So, if a firm cannot fully repay its debt, then the exact (partial) payments it makes to each of its creditors can affect the cash inflow back to itself. We naturally assume that the firms are interested in their financial well-being (utility) which is aligned with the amount of incoming payments they receive from the network. This defines a game among the firms, that can be seen as utility-maximizing agents who can strategize over their payments. We are the first to study financial network games that arise under a natural set of payment strategies called priority-proportional payments. We investigate the existence and (in)efficiency of equilibrium strategies, under different assumptions on how the firms' utility is defined, on the types of debt contracts allowed between the firms, and on the presence of other financial features that commonly arise in practice. Surprisingly, even if all firms' strategies are fixed, the existence of a unique payment profile is not guaranteed. So, we also investigate the existence and computation of valid payment profiles for fixed payment strategies.

cs.GT↗

arXiv2021

UniRE: A Unified Label Space for Entity Relation Extraction

Many joint entity relation extraction models setup two separated label spaces for the two sub-tasks (i.e., entity detection and relation classification). We argue that this setting may hinder the information interaction between entities and relations. In this work, we propose to eliminate the different treatment on the two sub-tasks' label spaces. The input of our model is a table containing all word pairs from a sentence. Entities and relations are represented by squares and rectangles in the table. We apply a unified classifier to predict each cell's label, which unifies the learning of two sub-tasks. For testing, an effective (yet fast) approximate decoder is proposed for finding squares and rectangles from tables. Experiments on three benchmarks (ACE04, ACE05, SciERC) show that, using only half the number of parameters, our model achieves competitive accuracy with the best extractor, and is faster.

cs.CL↗

arXiv2021

Probabilistic Graph Reasoning for Natural Proof Generation

In this paper, we investigate the problem of reasoning over natural language statements. Prior neural based approaches do not explicitly consider the inter-dependency among answers and their proofs. In this paper, we propose PRobr, a novel approach for joint answer prediction and proof generation. PRobr defines a joint probabilistic distribution over all possible proof graphs and answers via an induced graphical model. We then optimize the model using variational approximation on top of neural textual representation. Experiments on multiple datasets under diverse settings (fully supervised, few-shot and zero-shot evaluation) verify the effectiveness of PRobr, e.g., achieving 10%-30% improvement on QA accuracy in few/zero-shot evaluation. Our codes and models can be found at https://github.com/changzhisun/PRobr/.

cs.CL↗

arXiv2021

Short-Term Load Forecasting for Smart HomeAppliances with Sequence to Sequence Learning

Appliance-level load forecasting plays a critical role in residential energy management, besides having significant importance for ancillary services performed by the utilities. In this paper, we propose to use an LSTM-based sequence-to-sequence (seq2seq) learning model that can capture the load profiles of appliances. We use a real dataset collected fromfour residential buildings and compare our proposed schemewith three other techniques, namely VARMA, Dilated One Dimensional Convolutional Neural Network, and an LSTM model.The results show that the proposed LSTM-based seq2seq model outperforms other techniques in terms of prediction error in most cases.

eess.SP↗

arXiv2021

Embracing Uncertainty: Decoupling and De-bias for Robust Temporal Grounding

Temporal grounding aims to localize temporal boundaries within untrimmed videos by language queries, but it faces the challenge of two types of inevitable human uncertainties: query uncertainty and label uncertainty. The two uncertainties stem from human subjectivity, leading to limited generalization ability of temporal grounding. In this work, we propose a novel DeNet (Decoupling and De-bias) to embrace human uncertainty: Decoupling - We explicitly disentangle each query into a relation feature and a modified feature. The relation feature, which is mainly based on skeleton-like words (including nouns and verbs), aims to extract basic and consistent information in the presence of query uncertainty. Meanwhile, modified feature assigned with style-like words (including adjectives, adverbs, etc) represents the subjective information, and thus brings personalized predictions; De-bias - We propose a de-bias mechanism to generate diverse predictions, aim to alleviate the bias caused by single-style annotations in the presence of label uncertainty. Moreover, we put forward new multi-label metrics to diversify the performance evaluation. Extensive experiments show that our approach is more effective and robust than state-of-the-arts on Charades-STA and ActivityNet Captions datasets.

cs.CV↗

arXiv2021

RAN Resource Slicing in 5G Using Multi-Agent Correlated Q-Learning

5G is regarded as a revolutionary mobile network, which is expected to satisfy a vast number of novel services, ranging from remote health care to smart cities. However, heterogeneous Quality of Service (QoS) requirements of different services and limited spectrum make the radio resource allocation a challenging problem in 5G. In this paper, we propose a multi-agent reinforcement learning (MARL) method for radio resource slicing in 5G. We model each slice as an intelligent agent that competes for limited radio resources, and the correlated Q-learning is applied for inter-slice resource block (RB) allocation. The proposed correlated Q-learning based interslice RB allocation (COQRA) scheme is compared with Nash Q-learning (NQL), Latency-Reliability-Throughput Q-learning (LRTQ) methods, and the priority proportional fairness (PPF) algorithm. Our simulation results show that the proposed COQRA achieves 32.4% lower latency and 6.3% higher throughput when compared with LRTQ, and 5.8% lower latency and 5.9% higher throughput than NQL. Significantly higher throughput and lower packet drop rate (PDR) is observed in comparison to PPF.

cs.NI↗

arXiv2021

Improving Sign Language Translation with Monolingual Data by Sign Back-Translation

Despite existing pioneering works on sign language translation (SLT), there is a non-trivial obstacle, i.e., the limited quantity of parallel sign-text data. To tackle this parallel data bottleneck, we propose a sign back-translation (SignBT) approach, which incorporates massive spoken language texts into SLT training. With a text-to-gloss translation model, we first back-translate the monolingual text to its gloss sequence. Then, the paired sign sequence is generated by splicing pieces from an estimated gloss-to-sign bank at the feature level. Finally, the synthetic parallel data serves as a strong supplement for the end-to-end training of the encoder-decoder SLT framework. To promote the SLT research, we further contribute CSL-Daily, a large-scale continuous SLT dataset. It provides both spoken language translations and gloss-level annotations. The topic revolves around people's daily lives (e.g., travel, shopping, medical care), the most likely SLT application scenario. Extensive experimental results and analysis of SLT methods are reported on CSL-Daily. With the proposed sign back-translation method, we obtain a substantial improvement over previous state-of-the-art SLT methods.

cs.CV↗

arXiv2021

Glancing Transformer for Non-Autoregressive Neural Machine Translation

Recent work on non-autoregressive neural machine translation (NAT) aims at improving the efficiency by parallel decoding without sacrificing the quality. However, existing NAT methods are either inferior to Transformer or require multiple decoding passes, leading to reduced speedup. We propose the Glancing Language Model (GLM), a method to learn word interdependency for single-pass parallel generation models. With GLM, we develop Glancing Transformer (GLAT) for machine translation. With only single-pass parallel decoding, GLAT is able to generate high-quality translation with 8-15 times speedup. Experiments on multiple WMT language directions show that GLAT outperforms all previous single pass non-autoregressive methods, and is nearly comparable to Transformer, reducing the gap to 0.25-0.9 BLEU points.

cs.CL↗

arXiv2021

Compatibility-aware Heterogeneous Visual Search

We tackle the problem of visual search under resource constraints. Existing systems use the same embedding model to compute representations (embeddings) for the query and gallery images. Such systems inherently face a hard accuracy-efficiency trade-off: the embedding model needs to be large enough to ensure high accuracy, yet small enough to enable query-embedding computation on resource-constrained platforms. This trade-off could be mitigated if gallery embeddings are generated from a large model and query embeddings are extracted using a compact model. The key to building such a system is to ensure representation compatibility between the query and gallery models. In this paper, we address two forms of compatibility: One enforced by modifying the parameters of each model that computes the embeddings. The other by modifying the architectures that compute the embeddings, leading to compatibility-aware neural architecture search (CMP-NAS). We test CMP-NAS on challenging retrieval tasks for fashion images (DeepFashion2), and face images (IJB-C). Compared to ordinary (homogeneous) visual search using the largest embedding model (paragon), CMP-NAS achieves 80-fold and 23-fold cost reduction while maintaining accuracy within 0.3% and 1.6% of the paragon on DeepFashion2 and IJB-C respectively.

cs.CV↗

arXiv2021

Imposing equilibrium on measured 3-D stress fields using Hodge decomposition and FFT-based optimization

We present a methodology to impose micromechanical constraints, i.e. stress equilibrium at grain and sub-grain scale, to an arbitrary (non-equilibrated) voxelized stress field obtained, for example, by means of synchrotron X-ray diffraction techniques. The method consists in finding the equilibrated stress field closest (in $L^2$-norm sense) to the measured non-equilibrated stress field, via the solution of an optimization problem. The extraction of the divergence-free (equilibrated) part of a general (non-equilibrated) field is performed using the Hodge decomposition of a symmetric matrix field, which is the generalization of the Helmholtz decomposition of a vector field into the sum of an irrotational field and a solenoidal field. The combination of: a) the Euler-Lagrange equations that solve the optimization problem, and b) the Hodge decomposition, gives a differential expression that contains the bi-harmonic operator and two times the curl operator acting on the measured stress field. These high-order derivatives can be efficiently performed in Fourier space. The method is applied to filter the non-equilibrated parts of a synthetic piecewise constant stress fields with a known ground truth, and stress fields in Gum Metal, a beta-Ti-based alloy measured in-situ using Diffraction Contrast Tomography (DCT). In both cases, the largest corrections were obtained near grain boundaries.

cond-mat.mtrl-sci↗

arXiv2021

Multiscale modeling of materials: Computing, data science,uncertainty and goal-oriented optimization

The recent decades have seen various attempts at accelerating the process of developing materials targeted towards specific applications. The performance required for a particular application leads to the choice of a particular material system whose properties are optimized by manipulating its underlying microstructure through processing. The specific configuration of the structure is then designed by characterizing the material in detail, and using this characterization along with physical principles in system level simulations and optimization. These have been advanced by multiscale modeling of materials, high-throughput experimentations, materials data-bases, topology optimization and other ideas. Still, developing materials for extreme applications involving large deformation, high strain rates and high temperatures remains a challenge. This article reviews a number of recent methods that advance the goal of designing materials targeted by specific applications.

cond-mat.mtrl-sci↗

arXiv2021

"Listen, Understand and Translate": Triple Supervision Decouples End-to-end Speech-to-text Translation

An end-to-end speech-to-text translation (ST) takes audio in a source language and outputs the text in a target language. Existing methods are limited by the amount of parallel corpus. Can we build a system to fully utilize signals in a parallel ST corpus? We are inspired by human understanding system which is composed of auditory perception and cognitive processing. In this paper, we propose Listen-Understand-Translate, (LUT), a unified framework with triple supervision signals to decouple the end-to-end speech-to-text translation task. LUT is able to guide the acoustic encoder to extract as much information from the auditory input. In addition, LUT utilizes a pre-trained BERT model to enforce the upper encoder to produce as much semantic information as possible, without extra data. We perform experiments on a diverse set of speech translation benchmarks, including Librispeech English-French, IWSLT English-German and TED English-Chinese. Our results demonstrate LUT achieves the state-of-the-art performance, outperforming previous methods. The code is available at https://github.com/dqqcasia/st.

cs.CL↗