arXiv subjects

Hui Zhang

Publications and source records attributed to Hui Zhang.

At least 343 records · Page 19Linked to original sources

arXiv2022

Robust Speaker Extraction Network Based on Iterative Refined Adaptation

Speaker extraction aims to extract target speech signal from a multi-talker environment with interference speakers and surrounding noise, given the target speaker's reference information. Most speaker extraction systems achieve satisfactory performance on the premise that the test speakers have been encountered during training time. Such systems suffer from performance degradation given unseen target speakers and/or mismatched reference voiceprint information. In this paper we propose a novel strategy named Iterative Refined Adaptation (IRA) to improve the robustness and generalization capability of speaker extraction systems in the aforementioned scenarios. Given an initial speaker embedding encoded by an auxiliary network, the extraction network can obtain a latent representation of the target speaker, which is fed back to the auxiliary network to get a refined embedding to provide more accurate guidance for the extraction network. Experiments on WSJ0-2mix-extr and WHAM! dataset confirm the superior performance of the proposed method over the network without IRA in terms of SI-SDR and PESQ improvement.

eess.AS↗

arXiv2022

Enforcing continuous symmetries in physics-informed neural network for solving forward and inverse problems of partial differential equations

As a typical application of deep learning, physics-informed neural network (PINN) {has been} successfully used to find numerical solutions of partial differential equations (PDEs), but how to improve the limited accuracy is still a great challenge for PINN. In this work, we introduce a new method, symmetry-enhanced physics informed neural network (SPINN) where the invariant surface conditions induced by the Lie symmetries or non-classical symmetries of PDEs are embedded into the loss function in PINN, to improve the accuracy of PINN for solving the forward and inverse problems of PDEs. We test the effectiveness of SPINN for the forward problem via two groups of ten independent numerical experiments using different numbers of collocation points and neurons for the heat equation, Korteweg-de Vries (KdV) equation and potential Burgers {equations} respectively, and for the inverse problem by considering different layers and neurons as well as different training points for the Burgers equation in potential form. The numerical results show that SPINN performs better than PINN with fewer training points and simpler architecture of neural network. Furthermore, we discuss the computational overhead of SPINN in terms of the relative computational cost to PINN and show that the training time of SPINN has no obvious increases, even less than PINN for certain cases.

cs.LG↗

arXiv2022

Schwarz methods by domain truncation

Schwarz methods use a decomposition of the computational domain into subdomains and need to put boundary conditions on the subdomain boundaries. In domain truncation one restricts the unbounded domain to a bounded computational domain and also needs to put boundary conditions on the computational domain boundaries. It turns out to be fruitful to think of the domain decomposition in Schwarz methods as truncation of the domain onto subdomains. The first truly optimal Schwarz method that converges in a finite number of steps was proposed in 1994 and used precisely transparent boundary conditions as transmission conditions between subdomains. Approximating these transparent boundary conditions for fast convergence of Schwarz methods led to the development of optimized Schwarz methods -- a name that has become common for Schwarz methods based on domain truncation. Compared to classical Schwarz methods which use simple Dirichlet transmission conditions and have been successfully used in a wide range of applications, optimized Schwarz methods are much less well understood, mainly due to their more sophisticated transmission conditions. This present situation is the motivation for our survey: to give a comprehensive review and precise exploration of convergence behaviors of optimized Schwarz methods based on Fourier analysis taking into account the original boundary conditions, many subdomain decompositions and layered media. The transmission conditions we study include the lowest order absorbing conditions (Robin), and also more advanced perfectly matched layers (PML), both developed first for domain truncation.

math.NA↗

arXiv2022

Confluent Vandermonde with Arnoldi

In this note, we extend the Vandermonde with Arnoldi method recently advocated by P. D. Brubeck, Y. Nakatsukasa and L. N. Trefethen to dealing with the confluent Vandermonde matrix. To apply the Arnoldi process, it is critical to find a Krylov subspace which generates the column space of the confluent Vandermonde matrix. A theorem is established for such Krylov subspaces for any order derivatives. This enables us to compute the derivatives of high degree polynomials to high precision. It also makes many applications involving derivatives possible, as illustrated by numerical examples. We note that one of the approaches orthogonalizes only the function values and is equivalent to the formula given by P. D. Brubeck and L. N. Trefethen. The other approach orthogonalizes the Hermite data. About which approach is preferable to another, we made the comparison, and the result is problem dependent.

math.NA↗

arXiv2022

A Unified Understanding of Deep NLP Models for Text Classification

The rapid development of deep natural language processing (NLP) models for text classification has led to an urgent need for a unified understanding of these models proposed individually. Existing methods cannot meet the need for understanding different models in one framework due to the lack of a unified measure for explaining both low-level (e.g., words) and high-level (e.g., phrases) features. We have developed a visual analysis tool, DeepNLPVis, to enable a unified understanding of NLP models for text classification. The key idea is a mutual information-based measure, which provides quantitative explanations on how each layer of a model maintains the information of input words in a sample. We model the intra- and inter-word information at each layer measuring the importance of a word to the final prediction as well as the relationships between words, such as the formation of phrases. A multi-level visualization, which consists of a corpus-level, a sample-level, and a word-level visualization, supports the analysis from the overall training set to individual samples. Two case studies on classification tasks and comparison between models demonstrate that DeepNLPVis can help users effectively identify potential problems caused by samples and model architectures and then make informed improvements.

cs.CL↗

arXiv2022

ET White Paper: To Find the First Earth 2.0

We propose to develop a wide-field and ultra-high-precision photometric survey mission, temporarily named "Earth 2.0 (ET)". This mission is designed to measure, for the first time, the occurrence rate and the orbital distributions of Earth-sized planets. ET consists of seven 30cm telescopes, to be launched to the Earth-Sun's L2 point. Six of these are transit telescopes with a field of view of 500 square degrees. Staring in the direction that encompasses the original Kepler field for four continuous years, this monitoring will return tens of thousands of transiting planets, including the elusive Earth twins orbiting solar-type stars. The seventh telescope is a 30cm microlensing telescope that will monitor an area of 4 square degrees toward the galactic bulge. This, combined with simultaneous ground-based KMTNet observations, will measure masses for hundreds of long-period and free-floating planets. Together, the transit and the microlensing telescopes will revolutionize our understandings of terrestrial planets across a large swath of orbital distances and free space. In addition, the survey data will also facilitate studies in the fields of asteroseismology, Galactic archeology, time-domain sciences, and black holes in binaries.

astro-ph.IM↗

arXiv2022

Neural Network Decoders for Permutation Codes Correcting Different Errors

Permutation codes were extensively studied in order to correct different types of errors for the applications on power line communication and rank modulation for flash memory. In this paper, we introduce the neural network decoders for permutation codes to correct these errors with one-shot decoding, which treat the decoding as $n$ classification tasks for non-binary symbols for a code of length $n$. These are actually the first general decoders introduced to deal with any error type for these two applications. The performance of the decoders is evaluated by simulations with different error models.

cs.IT↗

arXiv2022

Localized semiclassical states for Hamiltonian elliptic systems in dimension two

In this paper, we consider the Hamiltonian elliptic system in dimension two\begin{equation}\label{1.5}\aligned \left\{ \begin{array}{lll} -ε^2Δu+V(x)u=g(v)\ & \text{in}\quad \mathbb{R}^2,\\ -ε^2Δv+V(x)v=f(u)\ & \text{in}\quad \mathbb{R}^2, \end{array}\right.\endaligned \end{equation} where $V\in C(\mathbb{R}^2)$ has local minimum points, and $f,g\in C^1(\mathbb{R})$ are assumed to be either superlinear or asymptotically linear at infinity and of subcritical exponential growth in the sense of Trudinger-Moser inequality. Under only a local condition on $V$, we obtain a family of semiclassical states concentrating around local minimum points of $V$. In addition, in the case that $f$ and $g$ are superlinear at infinity, the decay and positivity of semiclassical states are also given. The proof is based on a reduction method, variational methods and penalization techniques.

math.AP↗

arXiv2022

Nodal solutions for quasilinear Schrödinger equations with asymptotically 3-linear nonlinearity

In this paper, we are concerned with the quasilinear Schrödinger equation \begin{equation*} -Δu+V(x)u-uΔ(u^2)=g(u),\ \ x\in \mathbb{R}^{N}, \end{equation*} where $N\geq3$, $V$ is radially symmetric and nonnegative, and $g$ is asymptotically 3-linear at infinity. In the case of $\inf_{\mathbb{R}^N}V>0$, we show the existence of a least energy sign-changing solution with exactly one node, and for any integer $k>0$, there are a pair of sign-changing solutions with $k$ nodes. Moreover, in the case of $\inf_{\mathbb{R}^N}V=0$, the problem above admits a least energy sign-changing solution with exactly one node. The proof is based on variational methods. In particular, some new tricks and the method of sign-changing Nehari manifold depending on a suitable restricted set are introduced to overcome the difficulty resulting from the appearance of asymptotically 3-linear nonlinearities.

math.AP↗

arXiv2022

PaddleSpeech: An Easy-to-Use All-in-One Speech Toolkit

PaddleSpeech is an open-source all-in-one speech toolkit. It aims at facilitating the development and research of speech processing technologies by providing an easy-to-use command-line interface and a simple code structure. This paper describes the design philosophy and core architecture of PaddleSpeech to support several essential speech-to-text and text-to-speech tasks. PaddleSpeech achieves competitive or state-of-the-art performance on various speech datasets and implements the most popular methods. It also provides recipes and pretrained models to quickly reproduce the experimental results in this paper. PaddleSpeech is publicly avaiable at https://github.com/PaddlePaddle/PaddleSpeech.

eess.AS↗

arXiv2022

The TESS Grand Unified Hot Jupiter Survey. I. Ten TESS Planets

We report the discovery of ten short-period giant planets (TOI-2193A b, TOI-2207 b, TOI-2236 b, TOI-2421 b, TOI-2567 b, TOI-2570 b, TOI-3331 b, TOI-3540A b, TOI-3693 b, TOI-4137 b). All of the planets were identified as planet candidates based on periodic flux dips observed by NASA's Transiting Exoplanet Survey Satellite (TESS). The signals were confirmed to be from transiting planets using ground-based time-series photometry, high angular resolution imaging, and high-resolution spectroscopy coordinated with the TESS Follow-up Observing Program. The ten newly discovered planets orbit relatively bright F and G stars ($G < 12.5$,~$T_\mathrm{eff}$ between 4800 and 6200 K). The planets' orbital periods range from 2 to 10~days, and their masses range from 0.2 to 2.2 Jupiter masses. TOI-2421 b is notable for being a Saturn-mass planet and TOI-2567 b for being a ``sub-Saturn'', with masses of $0.322\pm 0.073$ and $0.195\pm 0.030$ Jupiter masses, respectively. In most cases, we have little information about the orbital eccentricities. Two exceptions are TOI-2207 b, which has an 8-day period and a detectably eccentric orbit ($e = 0.17\pm0.05$), and TOI-3693 b, a 9-day planet for which we can set an upper limit of $e < 0.052$. The ten planets described here are the first new planets resulting from an effort to use TESS data to unify and expand on the work of previous ground-based transit surveys in order to create a large and statistically useful sample of hot Jupiters.

astro-ph.EP↗

arXiv2022

3D Segmentation Guided Style-based Generative Adversarial Networks for PET Synthesis

Potential radioactive hazards in full-dose positron emission tomography (PET) imaging remain a concern, whereas the quality of low-dose images is never desirable for clinical use. So it is of great interest to translate low-dose PET images into full-dose. Previous studies based on deep learning methods usually directly extract hierarchical features for reconstruction. We notice that the importance of each feature is different and they should be weighted dissimilarly so that tiny information can be captured by the neural network. Furthermore, the synthesis on some regions of interest is important in some applications. Here we propose a novel segmentation guided style-based generative adversarial network (SGSGAN) for PET synthesis. (1) We put forward a style-based generator employing style modulation, which specifically controls the hierarchical features in the translation process, to generate images with more realistic textures. (2) We adopt a task-driven strategy that couples a segmentation task with a generative adversarial network (GAN) framework to improve the translation performance. Extensive experiments show the superiority of our overall framework in PET synthesis, especially on those regions of interest.

eess.IV↗

arXiv2022

A Roadmap for Big Model

With the rapid development of deep learning, training Big Models (BMs) for multiple downstream tasks becomes a popular paradigm. Researchers have achieved various outcomes in the construction of BMs and the BM application in many fields. At present, there is a lack of research work that sorts out the overall progress of BMs and guides the follow-up research. In this paper, we cover not only the BM technologies themselves but also the prerequisites for BM training and applications with BMs, dividing the BM review into four parts: Resource, Models, Key Technologies and Application. We introduce 16 specific BM-related topics in those four parts, they are Data, Knowledge, Computing System, Parallel Training System, Language Model, Vision Model, Multi-modal Model, Theory&Interpretability, Commonsense Reasoning, Reliability&Security, Governance, Evaluation, Machine Translation, Text Generation, Dialogue and Protein Research. In each topic, we summarize clearly the current studies and propose some future research directions. At the end of this paper, we conclude the further development of BMs in a more general view.

cs.LG↗

arXiv2022

HD 83443c: A highly eccentric giant planet on a 22-year orbit

We report the discovery of a highly eccentric long-period Jovian planet orbiting the hot-Jupiter host HD\,83443. By combining radial velocity data from four instruments (AAT/UCLES, Keck/HIRES, HARPS, Minerva-Australis) spanning more than two decades, we find evidence for a planet with m~sin~$i=1.35^{+0.07}_{-0.06}$\,\mj, moving on an orbit with $a=8.0\pm$0.8\,au and eccentricity $e=0.76\pm$0.05. We combine our radial velocity analysis with \textit{Gaia} eDR3 /\textit{Hipparcos} proper motion anomalies and derive a dynamical mass of $1.5^{+0.5}_{-0.2} M_{\rm Jup}$. We perform a detailed dynamical simulation that reveals locations of stability within the system that may harbor additional planets, including stable regions within the habitable zone of the host star. HD\,83443 is a rare example of a system hosting a hot Jupiter and an exterior planetary companion. The high eccentricity of HD\,83443c suggests that a scattering event may have sent the hot Jupiter to its close orbit while leaving the outer planet on a wide and eccentric path.

astro-ph.EP↗

arXiv2022

Doubly resonant photoacoustic spectroscopy: ultra-high sensitivity meets ultra-wide dynamic range

Photoacoustic spectroscopy (PAS) based gas sensors with high sensitivity, wide dynamic range, low cost, and small footprint are desirable across a broad range of applications in energy, environment, safety, and public health. However, most works have focused on either acoustic resonator to enhance acoustic wave or optical resonator to enhance optical wave. Herein, we develop a gas sensor based on doubly resonant PAS in which the acoustic and optical waves are simultaneously enhanced using combined optical and acoustic resonators in a centimeter-long configuration. Not only the lower detection limit is enhanced by the double standing waves, but also the upper detection limit is expanded due to the short resonators. As an example, we developed a sensor by detecting acetylene (C2H2), achieving a noise equivalent absorption of 5.7*10-13 cm-1 and a dynamic range of eight orders. Compared to the state-of-the-art PAS gas sensors, the developed sensor increases the sensitivity by two orders of magnitude and extends the dynamic range by three orders of magnitude. Besides, a laser-cavity-molecule locking strategy is proposed to provide additional flexibility of fast gas detection.

physics.app-ph↗

arXiv2022

Non-neglectable entropy effect on sintering of supported nanoparticles

Sintering refers to particle coalescence by heat, which has been known as a thermal phenomenon involving all aspects of natural science for centuries. It is particularly important in heterogeneous catalysis because normally sintering results in deactivation of the catalysts. In previous studies, the enthalpy contribution was considered to be dominant in sintering and the entropy effect is generally considered neglectable. However, we unambiguously demonstrate in this work that entropy could prevail over the enthalpy contribution to dominate the sintering behavior of supported nanoparticles (NPs) by designed experiments and improved theoretical framework. Using in situ Cs-corrected environmental scanning transmission electron microscopy and synchrotron-based ambient pressure X-ray photoelectron spectroscopy, we observe the unprecedent entropy-driven phenomenon that supported NPs reversibly redisperse upon heating and sinter upon cooling in three systems (Pd-CeO2, Cu-TiO2, Ag-TiO2). We quantitatively show that the configurational entropy of highly dispersed ad-atoms is large enough to reverse their sintering tendency at the elevated temperature. This work reshapes the basic understanding of sintering at the nanoscale and opens the door for various de-novo designs of thermodynamically stable nanocatalysts.

cond-mat.mtrl-sci↗

arXiv2022

Influence of moments of inertia on transverse wobbling mode in odd-mass nuclei

The reported transverse wobbling band in odd-mass $^{105}$Pd has been reinvestigated by the triaxial particle rotor model. Employing different parameter sets of moment of inertia (MOI), several calculated results could be in good agreement with the experimental data, which show distinct modes of rotational excitation, respectively. These modes are sensitive to the ratio between the MOI at intermediate and short axis. With the increase of this ratio, a wobble about the short axis of the total angular momentum is gradually changed to a wobble about the intermediate axis. In addition, it is exhibited that precession and tunneling are two aspects of the quantum wobbling motion. The tunneling aspect dominates in the yrare states of $^{105}$Pd. The present results in $^{105}$Pd show the complexity of the transverse wobbling mode.

nucl-th↗

arXiv2022

Coverage hole detection in WSN with force-directed algorithm and transfer learning

Coverage hole detection is an important research problem in wireless sensor network research community. However, distributed approaches proposed in recent years for coverage hole detection problem have high computational complexity. In this paper, we propose a novel approach for coverage hole detection in wireless sensor networks called FD-TL (Force-directed and Transfer-learning) which is based on layout generation capability of Force-directed Algorithms and image recognition power of Convolutional Neural Network with transfer learning. In contrast to existing approaches, the proposed approach is a pure topology-based approach since FD-TL can detect both triangular and non-triangular coverage holes from a wireless sensor network based on the input network topology without relying on the physical locations of the anchor nodes. In FD-TL, a Force-directed Algorithm is used to generate a series of possible layouts from a given input topology. Next, a Convolutional Neural Network is used to recognize potential coverage holes from the generated layouts. During the training phase, a transfer learning method is used to aid the recognition process. Experimental results show that FD-TL method can achieve 90% sensitivity and 96% specificity for coverage hole detection in wireless sensor networks.

cs.NI↗