arXiv subjects

Jian Wu

Publications and source records attributed to Jian Wu.

At least 289 records · Page 16Linked to original sources

arXiv2022

DANets: Deep Abstract Networks for Tabular Data Classification and Regression

Tabular data are ubiquitous in real world applications. Although many commonly-used neural components (e.g., convolution) and extensible neural networks (e.g., ResNet) have been developed by the machine learning community, few of them were effective for tabular data and few designs were adequately tailored for tabular data structures. In this paper, we propose a novel and flexible neural component for tabular data, called Abstract Layer (AbstLay), which learns to explicitly group correlative input features and generate higher-level features for semantics abstraction. Also, we design a structure re-parameterization method to compress the learned AbstLay, thus reducing the computational complexity by a clear margin in the reference phase. A special basic block is built using AbstLays, and we construct a family of Deep Abstract Networks (DANets) for tabular data classification and regression by stacking such blocks. In DANets, a special shortcut path is introduced to fetch information from raw tabular features, assisting feature interactions across different levels. Comprehensive experiments on seven real-world tabular datasets show that our AbstLay and DANets are effective for tabular data classification and regression, and the computational complexity is superior to competitive methods. Besides, we evaluate the performance gains of DANet as it goes deep, verifying the extendibility of our method. Our code is available at https://github.com/WhatAShot/DANet.

cs.LG↗

arXiv2022

Sound2Synth: Interpreting Sound via FM Synthesizer Parameters Estimation

Synthesizer is a type of electronic musical instrument that is now widely used in modern music production and sound design. Each parameters configuration of a synthesizer produces a unique timbre and can be viewed as a unique instrument. The problem of estimating a set of parameters configuration that best restore a sound timbre is an important yet complicated problem, i.e.: the synthesizer parameters estimation problem. We proposed a multi-modal deep-learning-based pipeline Sound2Synth, together with a network structure Prime-Dilated Convolution (PDC) specially designed to solve this problem. Our method achieved not only SOTA but also the first real-world applicable results on Dexed synthesizer, a popular FM synthesizer.

cs.SD↗

arXiv2022

Direct numerical simulations of incompressible multiphase electrohydrodynamic flow with single-phase transportation schemes

In the present study, two schemes named face discernment and flux correction are proposed to achieve single-phase transportation of free charge in multiphase electrohydrodynamic(EHD) problems. Many EHD phenomena occur between air and another liquid while the free charge can only be transported in the liquid phase through ohmic conduction and convection due to the poor conductivity of air. However, the charge may be leaked into the dielectric air during the simulation due to the asynchronous transportation between interface and free charge. To avoid this unphysical error, a face discernment method is designed to produce an accurate ohmic conduction of free charge by providing a superior physical properties distribution at the interface. Subsequently the flux correction method is developed to correct the advection flux of charge density to prevent ions crossing the interface. These two schemes are based on the Volume of Fluid (VOF) model and independent with the specific interface updating method. The performance of the proposed methods are carefully validated with several test cases. The algorithms are implemented as an OpenFOAM extension and are published as open source.

physics.flu-dyn↗

arXiv2022

Streaming Multi-Talker ASR with Token-Level Serialized Output Training

This paper proposes a token-level serialized output training (t-SOT), a novel framework for streaming multi-talker automatic speech recognition (ASR). Unlike existing streaming multi-talker ASR models using multiple output branches, the t-SOT model has only a single output branch that generates recognition tokens (e.g., words, subwords) of multiple speakers in chronological order based on their emission times. A special token that indicates the change of ``virtual'' output channels is introduced to keep track of the overlapping utterances. Compared to the prior streaming multi-talker ASR models, the t-SOT model has the advantages of less inference cost and a simpler model architecture. Moreover, in our experiments with LibriSpeechMix and LibriCSS datasets, the t-SOT-based transformer transducer model achieves the state-of-the-art word error rates by a significant margin to the prior results. For non-overlapping speech, the t-SOT model is on par with a single-talker ASR model in terms of both accuracy and computational cost, opening the door for deploying one model for both single- and multi-talker scenarios.

eess.AS↗

arXiv2022

Streaming Speaker-Attributed ASR with Token-Level Speaker Embeddings

This paper presents a streaming speaker-attributed automatic speech recognition (SA-ASR) model that can recognize ``who spoke what'' with low latency even when multiple people are speaking simultaneously. Our model is based on token-level serialized output training (t-SOT) which was recently proposed to transcribe multi-talker speech in a streaming fashion. To further recognize speaker identities, we propose an encoder-decoder based speaker embedding extractor that can estimate a speaker representation for each recognized token not only from non-overlapping speech but also from overlapping speech. The proposed speaker embedding, named t-vector, is extracted synchronously with the t-SOT ASR model, enabling joint execution of speaker identification (SID) or speaker diarization (SD) with the multi-talker transcription with low latency. We evaluate the proposed model for a joint task of ASR and SID/SD by using LibriSpeechMix and LibriCSS corpora. The proposed model achieves substantially better accuracy than a prior streaming model and shows comparable or sometimes even superior results to the state-of-the-art offline SA-ASR model.

eess.AS↗

arXiv2022

Why does Self-Supervised Learning for Speech Recognition Benefit Speaker Recognition?

Recently, self-supervised learning (SSL) has demonstrated strong performance in speaker recognition, even if the pre-training objective is designed for speech recognition. In this paper, we study which factor leads to the success of self-supervised learning on speaker-related tasks, e.g. speaker verification (SV), through a series of carefully designed experiments. Our empirical results on the Voxceleb-1 dataset suggest that the benefit of SSL to SV task is from a combination of mask speech prediction loss, data scale, and model size, while the SSL quantizer has a minor impact. We further employ the integrated gradients attribution method and loss landscape visualization to understand the effectiveness of self-supervised learning for speaker recognition performance.

cs.CL↗

arXiv2022

WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing

Self-supervised learning (SSL) achieves great success in speech recognition, while limited exploration has been attempted for other speech processing tasks. As speech signal contains multi-faceted information including speaker identity, paralinguistics, spoken content, etc., learning universal representations for all speech tasks is challenging. To tackle the problem, we propose a new pre-trained model, WavLM, to solve full-stack downstream speech tasks. WavLM jointly learns masked speech prediction and denoising in pre-training. By this means, WavLM does not only keep the speech content modeling capability by the masked speech prediction, but also improves the potential to non-ASR tasks by the speech denoising. In addition, WavLM employs gated relative position bias for the Transformer structure to better capture the sequence ordering of input speech. We also scale up the training dataset from 60k hours to 94k hours. WavLM Large achieves state-of-the-art performance on the SUPERB benchmark, and brings significant improvements for various speech processing tasks on their representative benchmarks. The code and pre-trained models are available at https://aka.ms/wavlm.

cs.CL↗

arXiv2022

Zeptosecond Angular Streak Camera

Time-resolved electronic processes on the attosecond scale have recently become experimentally accessible through the development of laser-based pump-probe interrogation techniques such as the attosecond streak camera, the reconstruction of attosecond beating by interference of two-photon transitions, and the attoclock. In this work, we demonstrate that by combining the concepts of the attosecond streak camera and the attoclock, time resolved processes down to the time scale of tens of zeptoseconds come into reach. Key to advancing to this remarkable level of time precision by this method termed the zeptosecond angular streak camera (ZASC) is its substantial intrinsic time-information redundancy. The ZASC results in a remarkably simple streaking trace, which is largely independent of the precise temporal structure of the streaking pulse, thereby bypassing the need for detailed characterization of the streaking field. Moreover, it is capable of retrieving information on the duration of the pump pulse. It is also capable of reaching attosecond-level precision in a single-shot mode that may be useful for free-electron-laser experiments. This concept promises to open pathways towards the chronoscopy of zeptosecond-level ultrafast processes.

physics.atom-ph↗

arXiv2022

Deploying self-supervised learning in the wild for hybrid automatic speech recognition

Self-supervised learning (SSL) methods have proven to be very successful in automatic speech recognition (ASR). These great improvements have been reported mostly based on highly curated datasets such as LibriSpeech for non-streaming End-to-End ASR models. However, the pivotal characteristics of SSL is to be utilized for any untranscribed audio data. In this paper, we provide a full exploration on how to utilize uncurated audio data in SSL from data pre-processing to deploying an streaming hybrid ASR model. More specifically, we present (1) the effect of Audio Event Detection (AED) model in data pre-processing pipeline (2) analysis on choosing optimizer and learning rate scheduling (3) comparison of recently developed contrastive losses, (4) comparison of various pre-training strategies such as utilization of in-domain versus out-domain pre-training data, monolingual versus multilingual pre-training data, multi-head multilingual SSL versus single-head multilingual SSL and supervised pre-training versus SSL. The experimental results show that SSL pre-training with in-domain uncurated data can achieve better performance in comparison to all the alternative out-domain pre-training strategies.

cs.SD↗

arXiv2022

What Can Machine Vision Do for Lymphatic Histopathology Image Analysis: A Comprehensive Review

In the past ten years, the computing power of machine vision (MV) has been continuously improved, and image analysis algorithms have developed rapidly. At the same time, histopathological slices can be stored as digital images. Therefore, MV algorithms can provide doctors with diagnostic references. In particular, the continuous improvement of deep learning algorithms has further improved the accuracy of MV in disease detection and diagnosis. This paper reviews the applications of image processing technology based on MV in lymphoma histopathological images in recent years, including segmentation, classification and detection. Finally, the current methods are analyzed, some more potential methods are proposed, and further prospects are made.

cs.CV↗

arXiv2022

SciEv: Finding Scientific Evidence Papers for Scientific News

In the past decade, many scientific news media that report scientific breakthroughs and discoveries emerged, bringing science and technology closer to the general public. However, not all scientific news article cites proper sources, such as original scientific papers. A portion of scientific news articles contain misinterpreted, exaggerated, or distorted information that deviates from facts asserted in the original papers. Manually identifying proper citations is laborious and costly. Therefore, it is necessary to automatically search for pertinent scientific papers that could be used as evidence for a given piece of scientific news. We propose a system called SciEv that searches for scientific evidence papers given a scientific news article. The system employs a 2-stage query paradigm with the first stage retrieving candidate papers and the second stage reranking them. The key feature of SciEv is it uses domain knowledge entities (DKEs) to find candidates in the first stage, which proved to be more effective than regular keyphrases. In the reranking stage, we explore different document representations for news articles and candidate papers. To evaluate our system, we compiled a pilot dataset consisting of 100 manually curated (news,paper) pairs from ScienceAlert and similar websites. To our best knowledge, this is the first dataset of this kind. Our experiments indicate that the transformer model performs the best for DKE extraction. The system achieves a P@1=50%, P@5=71%, and P@10=74% when it uses a TFIDF-based text representation. The transformer-based re-ranker achieves a comparable performance but costs twice as much time. We will collect more data and test the system for user experience.

cs.IR↗

arXiv2022

Ultra Fast Speech Separation Model with Teacher Student Learning

Transformer has been successfully applied to speech separation recently with its strong long-dependency modeling capacity using a self-attention mechanism. However, Transformer tends to have heavy run-time costs due to the deep encoder layers, which hinders its deployment on edge devices. A small Transformer model with fewer encoder layers is preferred for computational efficiency, but it is prone to performance degradation. In this paper, an ultra fast speech separation Transformer model is proposed to achieve both better performance and efficiency with teacher student learning (T-S learning). We introduce layer-wise T-S learning and objective shifting mechanisms to guide the small student model to learn intermediate representations from the large teacher model. Compared with the small Transformer model trained from scratch, the proposed T-S learning method reduces the word error rate (WER) by more than 5% for both multi-channel and single-channel speech separation on LibriCSS dataset. Utilizing more unlabeled speech data, our ultra fast speech separation models achieve more than 10% relative WER reduction.

eess.AS↗

arXiv2022

A State-of-the-art Survey of U-Net in Microscopic Image Analysis: from Simple Usage to Structure Mortification

Image analysis technology is used to solve the inadvertences of artificial traditional methods in disease, wastewater treatment, environmental change monitoring analysis and convolutional neural networks (CNN) play an important role in microscopic image analysis. An important step in detection, tracking, monitoring, feature extraction, modeling and analysis is image segmentation, in which U-Net has increasingly applied in microscopic image segmentation. This paper comprehensively reviews the development history of U-Net, and analyzes various research results of various segmentation methods since the emergence of U-Net and conducts a comprehensive review of related papers. First, this paper has summarized the improved methods of U-Net and then listed the existing significance of image segmentation techniques and their improvements that has introduced over the years. Finally, focusing on the different improvement strategies of U-Net in different papers, the related work of each application target is reviewed according to detailed technical categories to facilitate future research. Researchers can clearly see the dynamics of transmission of technological development and keep up with future trends in this interdisciplinary field.

eess.IV↗

arXiv2022

Electrocardio Panorama: Synthesizing New ECG Views with Self-supervision

Multi-lead electrocardiogram (ECG) provides clinical information of heartbeats from several fixed viewpoints determined by the lead positioning. However, it is often not satisfactory to visualize ECG signals in these fixed and limited views, as some clinically useful information is represented only from a few specific ECG viewpoints. For the first time, we propose a new concept, Electrocardio Panorama, which allows visualizing ECG signals from any queried viewpoints. To build Electrocardio Panorama, we assume that an underlying electrocardio field exists, representing locations, magnitudes, and directions of ECG signals. We present a Neural electrocardio field Network (Nef-Net), which first predicts the electrocardio field representation by using a sparse set of one or few input ECG views and then synthesizes Electrocardio Panorama based on the predicted representations. Specially, to better disentangle electrocardio field information from viewpoint biases, a new Angular Encoding is proposed to process viewpoint angles. Also, we propose a self-supervised learning approach called Standin Learning, which helps model the electrocardio field without direct supervision. Further, with very few modifications, Nef-Net can also synthesize ECG signals from scratch. Experiments verify that our Nef-Net performs well on Electrocardio Panorama synthesis, and outperforms the previous work on the auxiliary tasks (ECG view transformation and ECG synthesis from scratch). The codes and the division labels of cardiac cycles and ECG deflections on Tianchi ECG and PTB datasets are available at https://github.com/WhatAShot/Electrocardio-Panorama.

eess.SP↗

arXiv2022

The effect of $f$-$c$ hybridization on the $γ\rightarrowα$ phase transition of cerium studied by lanthanum doping

The hybridization between the localized 4$f$ level ($f$) with conduction ($c$) states in $γ$-Ce upon cooling has been previously revealed in single crystalline thin films experimentally and theoretically, whereas its influence on the $γ\rightarrowα$ phase transition was not explicitly verified, due to the fact that the phase transition happened in the bulk-layer, leaving the surface in the $γ$ phase. Here in our work, we circumvent this issue by investigating the effect of alloying addition of La on Ce, by means of crystal structure, electronic transport and ARPES measurements, together with a phenomenological periodic Anderson model and a modified Anderson impurity model. Our current researches indicate that the weakening of $f$-$c$ hybridization is the major factor in the suppression of $γ\rightarrowα$ phase transition by La doping. The consistency of our results with the effects of other rare earth and actinide alloying additions on the $γ\rightarrowα$ phase transition of Ce is also discussed. Our work demonstrates the importance of the interaction of $f$ and $c$ electrons in understanding the unconventional phase transition in Ce, which is intuitive for further researches on other rare earth and actinide metals and alloys with similar phase transition behaviors.

cond-mat.str-el↗

arXiv2022

Large-Scale 3D Semantic Reconstruction for Automated Driving Vehicles with Adaptive Truncated Signed Distance Function

The Large-scale 3D reconstruction, texturing and semantic mapping are nowadays widely used for automated driving vehicles, virtual reality and automatic data generation. However, most approaches are developed for RGB-D cameras with colored dense point clouds and not suitable for large-scale outdoor environments using sparse LiDAR point clouds. Since a 3D surface can be usually observed from multiple camera images with different view poses, an optimal image patch selection for the texturing and an optimal semantic class estimation for the semantic mapping are still challenging. To address these problems, we propose a novel 3D reconstruction, texturing and semantic mapping system using LiDAR and camera sensors. An Adaptive Truncated Signed Distance Function is introduced to describe surfaces implicitly, which can deal with different LiDAR point sparsities and improve model quality. The from this implicit function extracted triangle mesh map is then textured from a series of registered camera images by applying an optimal image patch selection strategy. Besides that, a Markov Random Field-based data fusion approach is proposed to estimate the optimal semantic class for each triangle mesh. Our approach is evaluated on a synthetic dataset, the KITTI dataset and a dataset recorded with our experimental vehicle. The results show that the 3D models generated using our approach are more accurate in comparison to using other state-of-the-art approaches. The texturing and semantic mapping achieve also very promising results.

cs.CV↗

arXiv2022

Maximizing Audio Event Detection Model Performance on Small Datasets Through Knowledge Transfer, Data Augmentation, And Pretraining: An Ablation Study

An Xception model reaches state-of-the-art (SOTA) accuracy on the ESC-50 dataset for audio event detection through knowledge transfer from ImageNet weights, pretraining on AudioSet, and an on-the-fly data augmentation pipeline. This paper presents an ablation study that analyzes which components contribute to the boost in performance and training time. A smaller Xception model is also presented which nears SOTA performance with almost a third of the parameters.

cs.SD↗

arXiv2022

Explanation of nearby SNRs for primary electron excess and proton spectral bump

Several groups have reported a possible excess of primary electrons at high energies with the joint fit of the positron fraction and total electron/positron spectra. With the latest release of high-precision electron/positron spectra measured by AMS-02, we further confirm this excess by fitting $ΔΦ$ $\rm(i.e., Φ_{e^-}-Φ_{e^+})$ data in this work. Then we investigate the contribution of a single nearby supernova remnant to the primary electron excess and find that Monogem can reasonably account for this excess. Moreover, we predict that the electron spectrum may harden again at a few TeVs due to Vela's contribution. DAMPE, which can accurately measure electrons at TeV scale, is expected to provide the robust test of this new spectral feature in the near future. Finally, we fit the proton spectrum data of DAMPE with Monogem or Loop I. We find that both the primary electron excess and the proton spectral bump could be mainly generated by Monogem.

astro-ph.HE↗