arXiv subjects

Chen Zhang

Publications and source records attributed to Chen Zhang.

At least 505 records · Page 28Linked to original sources

arXiv2022

L3DAS22 Challenge: Learning 3D Audio Sources in a Real Office Environment

The L3DAS22 Challenge is aimed at encouraging the development of machine learning strategies for 3D speech enhancement and 3D sound localization and detection in office-like environments. This challenge improves and extends the tasks of the L3DAS21 edition. We generated a new dataset, which maintains the same general characteristics of L3DAS21 datasets, but with an extended number of data points and adding constrains that improve the baseline model's efficiency and overcome the major difficulties encountered by the participants of the previous challenge. We updated the baseline model of Task 1, using the architecture that ranked first in the previous challenge edition. We wrote a new supporting API, improving its clarity and ease-of-use. In the end, we present and discuss the results submitted by all participants. L3DAS22 Challenge website: www.l3das.com/icassp2022.

eess.AS↗

arXiv2022

SQuant: On-the-Fly Data-Free Quantization via Diagonal Hessian Approximation

Quantization of deep neural networks (DNN) has been proven effective for compressing and accelerating DNN models. Data-free quantization (DFQ) is a promising approach without the original datasets under privacy-sensitive and confidential scenarios. However, current DFQ solutions degrade accuracy, need synthetic data to calibrate networks, and are time-consuming and costly. This paper proposes an on-the-fly DFQ framework with sub-second quantization time, called SQuant, which can quantize networks on inference-only devices with low computation and memory requirements. With the theoretical analysis of the second-order information of DNN task loss, we decompose and approximate the Hessian-based optimization objective into three diagonal sub-items, which have different areas corresponding to three dimensions of weight tensor: element-wise, kernel-wise, and output channel-wise. Then, we progressively compose sub-items and propose a novel data-free optimization objective in the discrete domain, minimizing Constrained Absolute Sum of Error (or CASE in short), which surprisingly does not need any dataset and is even not aware of network architecture. We also design an efficient algorithm without back-propagation to further reduce the computation complexity of the objective solver. Finally, without fine-tuning and synthetic datasets, SQuant accelerates the data-free quantization process to a sub-second level with >30% accuracy improvement over the existing data-free post-training quantization works, with the evaluated models under 4-bit quantization. We have open-sourced the SQuant framework at https://github.com/clevercool/SQuant.

cs.LG↗

arXiv2022

A two-step backward compatible fullband speech enhancement system

Speech enhancement methods based on deep learning have surpassed traditional methods. While many of these new approaches are operating on the wideband (16kHz) sample rate, a new fullband (48kHz) speech enhancement system is proposed in this paper. Compared to the existing fullband systems that utilizes perceptually motivated features to train the fullband speech enhancement using a single network structure, the proposed system is a two-step system ensuring good fullband speech enhancement quality while backward compatible to the existing wideband systems.

eess.AS↗

arXiv2022

Ultrafast Optical Spectroscopy Evidence of Pseudogap and Electron-Phonon Coupling in an Iron-Based Superconductor KCa$_2$Fe$_4$As$_4$F$_2$

We use ultrafast optical spectroscopy to study the nonequilibrium quasiparticle relaxation dynamics of the iron-based superconductor KCa$_2$Fe$_4$As$_4$F$_2$ with $T_c=33.5$ K. Our results reveal a possible pseudogap ($Δ_{PG}$ = 2.4 $\pm$ 0.1 meV) below $T^*\approx 50$ K but prior to the opening of a superconducting gap ($Δ_{SC}$(0) $\approx$ 4.3 $\pm$ 0.1 meV). Measurements under high pump fluence real two distinct, coherent phonon oscillations with 1.95 and 5.51 THz frequencies, respectively. The high-frequency $A_{1g}$(2) mode corresponds to the $c-$axis polarized vibrations of FeAs planes with a nominal electron-phonon coupling constant $λ_{A_{1g}(2)}$ = 0.194 $\pm$ 0.02. Our findings suggest that the pseudogap is likely a precursor of superconductivity, and the electron-phonon coupling may play an essential role in the superconducting pairing in KCa$_2$Fe$_4$As$_4$F$_2$.

cond-mat.supr-con↗

arXiv2022

MDD-Eval: Self-Training on Augmented Data for Multi-Domain Dialogue Evaluation

Chatbots are designed to carry out human-like conversations across different domains, such as general chit-chat, knowledge exchange, and persona-grounded conversations. To measure the quality of such conversational agents, a dialogue evaluator is expected to conduct assessment across domains as well. However, most of the state-of-the-art automatic dialogue evaluation metrics (ADMs) are not designed for multi-domain evaluation. We are motivated to design a general and robust framework, MDD-Eval, to address the problem. Specifically, we first train a teacher evaluator with human-annotated data to acquire a rating skill to tell good dialogue responses from bad ones in a particular domain and then, adopt a self-training strategy to train a new evaluator with teacher-annotated multi-domain data, that helps the new evaluator to generalize across multiple domains. MDD-Eval is extensively assessed on six dialogue evaluation benchmarks. Empirical results show that the MDD-Eval framework achieves a strong performance with an absolute improvement of 7% over the state-of-the-art ADMs in terms of mean Spearman correlation scores across all the evaluation benchmarks.

cs.CL↗

arXiv2022

Dark Confinement and Chiral Phase Transitions: Gravitational Waves vs Matter Representations

We study the gravitational-wave signal stemming from strongly coupled models featuring both, dark chiral and confinement phase transitions. We therefore identify strongly coupled theories that can feature a first-order phase transition. Employing the Polyakov-Nambu-Jona-Lasinio model, we focus our attention on SU(3) Yang-Mills theories featuring fermions in fundamental, adjoint, and two-index symmetric representations. We discover that for the gravitational-wave signals analysis, there are significant differences between the various representations. Interestingly we also observe that the two-index symmetric representation leads to the strongest first-order phase transition and therefore to a higher chance of being detected by the Big Bang Observer experiment. Our study of the confinement and chiral phase transitions is further applicable to extensions of the Standard Model featuring composite dynamics.

hep-ph↗

arXiv2021

Stellar Structure and Stability of Charged Interacting Quark Stars and Their Scaling Behaviour

We explore the stellar structure and radial stability of charged quark stars composed of interacting quark matter (IQM) in three classes of commonly used charge models. We adopt a general parametrization of IQM equation of state that includes the corrections from perturbative QCD, color superconductivity, and the strange quark mass into one parameter $λ$, or one dimensionless parameter $\barλ=λ^2/(4B_{\rm eff})$ after being rescaled with the effective bag constant $B_{\rm eff}$. We find that increasing charge tends to increase the mass and radius profiles, and enlarges the separation size in mass between the maximum mass point and the point where zero eigenfrequencies $ω^2_0=0$ of the fundamental radial oscillation mode occur. The sign of the separation in central density depends on the charge model; this separation also has a dependence on $λ$ such that increasing $λ$ (which can occur for either large color superconductivity or small strange quark mass) tends to decrease this separation size for the first and third classes of charge models monotonically. Moreover, for the second and third classes of charge models, we manage to numerically and analytically identify a new kind of stellar structure with a zero central pressure but a finite radius and mass. All the calculations and analysis are performed in a general dimensionless rescaling approach so that the results are independent of explicit values of dimensional parameters.

astro-ph.HE↗

arXiv2021

Automatic Evaluation and Moderation of Open-domain Dialogue Systems

The development of Open-Domain Dialogue Systems (ODS)is a trending topic due to the large number of research challenges, large societal and business impact, and advances in the underlying technology. However, the development of these kinds of systems requires two important characteristics:1) automatic evaluation mechanisms that show high correlations with human judgements across multiple dialogue evaluation aspects (with explainable features for providing constructive and explicit feedback on the quality of generative models' responses for quick development and deployment)and 2) mechanisms that can help to control chatbot responses,while avoiding toxicity and employing intelligent ways to handle toxic user comments and keeping interaction flow and engagement. This track at the 10th Dialogue System Technology Challenge (DSTC10) is part of the ongoing effort to promote scalable and toxic-free ODS. This paper describes the datasets and baselines provided to participants, as well as submission evaluation results for each of the two proposed subtasks.

cs.CL↗

arXiv2021

Boosting Mobile CNN Inference through Semantic Memory

Human brains are known to be capable of speeding up visual recognition of repeatedly presented objects through faster memory encoding and accessing procedures on activated neurons. For the first time, we borrow and distill such a capability into a semantic memory design, namely SMTM, to improve on-device CNN inference. SMTM employs a hierarchical memory architecture to leverage the long-tail distribution of objects of interest, and further incorporates several novel techniques to put it into effects: (1) it encodes high-dimensional feature maps into low-dimensional, semantic vectors for low-cost yet accurate cache and lookup; (2) it uses a novel metric in determining the exit timing considering different layers' inherent characteristics; (3) it adaptively adjusts the cache size and semantic vectors to fit the scene dynamics. SMTM is prototyped on commodity CNN engine and runs on both mobile CPU and GPU. Extensive experiments on large-scale datasets and models show that SMTM can significantly speed up the model inference over standard approach (up to 2X) and prior cache designs (up to 1.5X), with acceptable accuracy loss.

cs.CV↗

arXiv2021

A Practical Method for Automated Modeling and Parametric Stability Analysis of VSC with Periodical Steady State

Linear Time Periodic (LTP) framework-based analysis of Voltage Source Converters (VSCs) is becoming popular, a driven factor is that many of the existing VSC applications inevitably exhibit the periodic steady-state (PSS), e.g., VSCs with unbalanced grid connections and the operation of a modular multilevel converter (MMC). In these studies, acquisition of the VSC's PSS conditions is a necessary precondition for proper linearization and stability analysis, and the efficiency of this process is particularly important for parametric studies. To this end, this work develops a computational method for automating the LTP analyses of VSCs with an integrated PSS solver. The core of the method lies in a unified frequency-domain iteration process that is developed by applying the generalized averaging principle. Given this, modeling, stability analysis, as well as the solution of PSS conditions can be unified in one process. The algorithmic implementation of the method in MATLAB is elaborated. Application of the obtained tool in impedance generation and parametric stability test is presented with an unbalanced grid-tied VSC as the exemplification. Finally, PSCAD/EMTDC simulations further consolidate the validity of the results.

eess.SY↗

arXiv2021

Quantum-assisted Distortion-free audio signal sensing

Quantum sensors are keeping the cutting-edge sensitivities in metrology. However, for high-sensitive measurements of arbitrary signals, limitations in linear dynamic range could introduce distortions when sensing the frequency, magnitude and phase of unknown signals. Here, we overcome these limitations with advanced sensing protocol that combines quantum phase-sensitive detection with heterodyne readout. We present theoretical and experimental investigations using nitrogen-vacancy centers in diamond, showing the ability to sense radio signals with a 98 dB linear dynamic range, a 31 pT/Hz$^{1/2}$ sensitivity, and arbitrary frequency resolution. Further, we perform the quantum-assisted distortion-free audio signal (melody, speech) sensing with high fidelity. The methods developed here could broaden the horizon for quantum sensors towards applications in telecommunication, where high-fidelity and low-distortion at multiple frequency bands within small sensing volumes are required.

quant-ph↗

arXiv2021

Gravitational wave echoes from interacting quark stars

We show that interacting quark stars (IQSs) composed of interacting quark matter (IQM), including the strong interaction effects such as perturbative QCD corrections and color superconductivity, can be compact enough to feature a photon sphere that is essential to the signature of gravitational wave echoes. We utilize an IQM equation of state unifying all interacting phases by a simple reparametrization and rescaling, through which we manage to maximally reduce the number of degrees of freedom into one dimensionless parameter $\barλ$ that characterizes the relative size of strong interaction effects. It turns out that gravitational wave echoes are possible for IQSs with $\barλ\gtrsim10$ at large center pressure. Rescaling the dimension back, we illustrate its implication on the dimensional parameter space of effective bag constant $B_{\rm eff}$ and the superconducting gap $Δ$ with variations of the perturbative QCD parameter $a_4$ and the strange quark mass $m_s$. We calculate the rescaled GW echo frequencies $\bar{f}_\text{echo}$ associated with IQSs, from which we obtain a simple scaling relation for the minimal echo frequency $f_\text{echo}^{\rm min}\approx 5.76 {\sqrt{B_{\rm eff}/\text{(100 MeV)}^4}} \,\,\, \rm kHz$ at the large $\barλ$ limit.

hep-ph↗

arXiv2021

Investigating the Impact of Pre-trained Language Models on Dialog Evaluation

Recently, there is a surge of interest in applying pre-trained language models (Pr-LM) in automatic open-domain dialog evaluation. Pr-LMs offer a promising direction for addressing the multi-domain evaluation challenge. Yet, the impact of different Pr-LMs on the performance of automatic metrics is not well-understood. This paper examines 8 different Pr-LMs and studies their impact on three typical automatic dialog evaluation metrics across three different dialog evaluation benchmarks. Specifically, we analyze how the choice of Pr-LMs affects the performance of automatic metrics. Extensive correlation analyses on each of the metrics are performed to assess the effects of different Pr-LMs along various axes, including pre-training objectives, dialog evaluation criteria, model size, and cross-dataset robustness. This study serves as the first comprehensive assessment of the effects of different Pr-LMs on automatic dialog evaluation.

cs.CL↗

arXiv2021

OSOA: One-Shot Online Adaptation of Deep Generative Models for Lossless Compression

Explicit deep generative models (DGMs), e.g., VAEs and Normalizing Flows, have shown to offer an effective data modelling alternative for lossless compression. However, DGMs themselves normally require large storage space and thus contaminate the advantage brought by accurate data density estimation. To eliminate the requirement of saving separate models for different target datasets, we propose a novel setting that starts from a pretrained deep generative model and compresses the data batches while adapting the model with a dynamical system for only one epoch. We formalise this setting as that of One-Shot Online Adaptation (OSOA) of DGMs for lossless compression and propose a vanilla algorithm under this setting. Experimental results show that vanilla OSOA can save significant time versus training bespoke models and space versus using one model for all targets. With the same adaptation step number or adaptation time, it is shown vanilla OSOA can exhibit better space efficiency, e.g., $47\%$ less space, than fine-tuning the pretrained model and saving the fine-tuned model. Moreover, we showcase the potential of OSOA and motivate more sophisticated OSOA algorithms by showing further space or time efficiency with multiple updates per batch and early stopping.

cs.LG↗

arXiv2021

iVPF: Numerical Invertible Volume Preserving Flow for Efficient Lossless Compression

It is nontrivial to store rapidly growing big data nowadays, which demands high-performance lossless compression techniques. Likelihood-based generative models have witnessed their success on lossless compression, where flow based models are desirable in allowing exact data likelihood optimisation with bijective mappings. However, common continuous flows are in contradiction with the discreteness of coding schemes, which requires either 1) imposing strict constraints on flow models that degrades the performance or 2) coding numerous bijective mapping errors which reduces the efficiency. In this paper, we investigate volume preserving flows for lossless compression and show that a bijective mapping without error is possible. We propose Numerical Invertible Volume Preserving Flow (iVPF) which is derived from the general volume preserving flows. By introducing novel computation algorithms on flow models, an exact bijective mapping is achieved without any numerical error. We also propose a lossless compression algorithm based on iVPF. Experiments on various datasets show that the algorithm based on iVPF achieves state-of-the-art compression ratio over lightweight compression algorithms.

cs.LG↗

arXiv2021

Conditional Variational Autoencoder for Learned Image Reconstruction

Learned image reconstruction techniques using deep neural networks have recently gained popularity, and have delivered promising empirical results. However, most approaches focus on one single recovery for each observation, and thus neglect the uncertainty information. In this work, we develop a novel computational framework that approximates the posterior distribution of the unknown image at each query observation. The proposed framework is very flexible: It handles implicit noise models and priors, it incorporates the data formation process (i.e., the forward operator), and the learned reconstructive properties are transferable between different datasets. Once the network is trained using the conditional variational autoencoder loss, it provides a computationally efficient sampler for the approximate posterior distribution via feed-forward propagation, and the summarizing statistics of the generated samples are used for both point-estimation and uncertainty quantification. We illustrate the proposed framework with extensive numerical experiments on positron emission tomography (with both moderate and low count levels) showing that the framework generates high-quality samples when compared with state-of-the-art methods.

cs.CV↗

arXiv2021

Transient Synchronization Stability Analysis of Wind Farms with MMC-HVDC Integration Under Offshore AC Grid Fault

The MMC-HVDC connected offshore wind farms (OWFs) could suffer short circuit fault (SCF), whereas their transient stability is not well analysed. In this paper, the mechanism of the loss of synchronization (LOS) of this system is analysed considering the whole system state from the fault-on to the post-fault, and the discussion on fault type and fault clearance is addressed as well. A stability index is proposed to quantify the transient synchronization stability (TSS) of the system, which is capable to not only estimate whether the wind turbine generators (WTGs) be able to get resynchronized with the offshore MMC after the fault is cleared, but also to evaluate the performance of stability improving methods as well. Finally, a scenario of six cases is tested on the PSCAD/EMTDC simulation platform, where the performances of four existing stability improving methods are thoroughly compared via both numerical simulation and the proposed stability index.

eess.SY↗

arXiv2021

Twist-diameter coupling drives DNA twist changes by salt and temperature

DNA deformations play crucial roles in many biological processes and material applications. During DNA deformation, DNA structural parameters often exhibit non-trivial and counterintuitive couplings, such as the twist-stretch and twist-bending couplings. Here, we reveal an unexpectedly strong negative twist-diameter coupling through the synergy of magnetic-tweezers experiments, atomistic molecular dynamics simulations, and theoretical calculations. In experiments, the DNA twist angle always increases with the concentration of NaCl, KCl, or RbCl. Our simulations quantitatively reproduce salt-induced twist changes and reveal the underlying physical mechanism: the reduction of DNA diameter under a high salt concentration leads to the increase in DNA twist angle through a strong negative twist-diameter coupling. The twist-diameter coupling is mediated by two dihedral angles in DNA structure and the coupling constant is 4.5 kBT/(deg nm) for one base-pair. Based on this coupling constant, we predict the temperature-dependence of DNA twist -0.0102 deg/K per bp, which agrees with our and previous experimental results. Our analysis suggests that the twist-diameter coupling is a common driving force for salt- and temperature-induced DNA twist changes.

physics.bio-ph↗