arXiv subjects

Zhenyu Zhang

Publications and source records attributed to Zhenyu Zhang.

At least 271 records · Page 15Linked to original sources

arXiv2021

MediumVC: Any-to-any voice conversion using synthetic specific-speaker speeches as intermedium features

To realize any-to-any (A2A) voice conversion (VC), most methods are to perform symmetric self-supervised reconstruction tasks (Xi to Xi), which usually results in inefficient performances due to inadequate feature decoupling, especially for unseen speakers. We propose a two-stage reconstruction task (Xi to Yi to Xi) using synthetic specific-speaker speeches as intermedium features, where A2A VC is divided into two stages: any-to-one (A2O) and one-to-Any (O2A). In the A2O stage, we propose a new A2O method: SingleVC, by employing a noval data augment strategy(pitch-shifted and duration-remained, PSDR) to accomplish Xi to Yi. In the O2A stage, MediumVC is proposed based on pre-trained SingleVC to conduct Yi to Xi. Through such asymmetrical reconstruction tasks (Xi to Yi in SingleVC and Yi to Xi in MediumVC), the models are to capture robust disentangled features purposefully. Experiments indicate MediumVC can enhance the similarity of converted speeches while maintaining a high degree of naturalness.

eess.AS↗

arXiv2021

Magnetic moment preservation and emergent Kondo resonance of Co-phthalocyanine on semimetallic Sb(111)

Magnetic molecules on surfaces have been widely investigated to reveal delicate interfacial couplings and for potential technological applications. In these endeavors, one prevailing challenge is how to preserve or recover the molecular spins, especially on highly metallic substrates that can readily quench the magnetic moments of the ad-molecules. Here we use scanning tunneling microscopy/spectroscopy to exploit the semimetallic nature of antimony and observe, surprisingly yet pleasantly, that the spin of Co-phthalocyanine is well preserved on Sb(111), as unambiguously evidenced by the emergent strong Kondo resonance across the molecule. Our first-principles calculations further confirm that the optimal density of states near the Fermi level of the semimetal is a decisive factor, weakening the overall interfacial coupling, while still ensuring sufficiently effective electron-spin scattering in the many-body system. Beyond isolated ad-molecules, we discover that each of the magnetic moments in a molecular dimer or a densely packed island is distinctly preserved as well, rendering such molecular magnets immense potentials for ultra-high density memory devices.

cond-mat.mtrl-sci↗

arXiv2021

DialogueBERT: A Self-Supervised Learning based Dialogue Pre-training Encoder

With the rapid development of artificial intelligence, conversational bots have became prevalent in mainstream E-commerce platforms, which can provide convenient customer service timely. To satisfy the user, the conversational bots need to understand the user's intention, detect the user's emotion, and extract the key entities from the conversational utterances. However, understanding dialogues is regarded as a very challenging task. Different from common language understanding, utterances in dialogues appear alternately from different roles and are usually organized as hierarchical structures. To facilitate the understanding of dialogues, in this paper, we propose a novel contextual dialogue encoder (i.e. DialogueBERT) based on the popular pre-trained language model BERT. Five self-supervised learning pre-training tasks are devised for learning the particularity of dialouge utterances. Four different input embeddings are integrated to catch the relationship between utterances, including turn embedding, role embedding, token embedding and position embedding. DialogueBERT was pre-trained with 70 million dialogues in real scenario, and then fine-tuned in three different downstream dialogue understanding tasks. Experimental results show that DialogueBERT achieves exciting results with 88.63% accuracy for intent recognition, 94.25% accuracy for emotion recognition and 97.04% F1 score for named entity recognition, which outperforms several strong baselines by a large margin.

cs.CL↗

arXiv2021

Electronic structure and signature of Tomonaga-Luttinger liquid state in epitaxial CoSb$_{1-x}$ nanoribbons

Recently, monolayer CoSb/SrTiO$_3$ has been proposed as a candidate harboring interfacial superconductivity in analogy with monolayer FeSe/SrTiO$_3$. Experimentally, while the CoSb-based compounds manifesting as nanowires and thin films have been realized on SrTiO$_3$ substrates, serving as a rich playground, their electronic structures are still unknown and yet to be resolved. Here, we have fabricated CoSb$_{1-x}$ nanoribbons with quasi-one-dimensional stripes on SrTiO$_3$(001) substrates using molecular beam epitaxy, and investigated the electronic structure by in situ angle-resolved photoemission spectroscopy. Straight Fermi surfaces without lateral dispersions are observed. CoSb$_{1-x}$/SrTiO$_3$ is slightly hole doped, where the interfacial charge transfer is opposite to that in monolayer FeSe/SrTiO$_3$. The spectral weight near Fermi level exhibits power-law-like suppression and obeys a universal temperature scaling, serving as the signature of Tomonaga-Luttinger liquid (TLL) state. The obtained TLL parameter of $\sim$0.21 shows the underlying strong correlations. Our results not only suggest CoSb$_{1-x}$ nanoribbon as a representative TLL system, but also provide clues for further investigations on the CoSb-related interface.

cond-mat.str-el↗

arXiv2021

Regularizing Nighttime Weirdness: Efficient Self-supervised Monocular Depth Estimation in the Dark

Monocular depth estimation aims at predicting depth from a single image or video. Recently, self-supervised methods draw much attention since they are free of depth annotations and achieve impressive performance on several daytime benchmarks. However, they produce weird outputs in more challenging nighttime scenarios because of low visibility and varying illuminations, which bring weak textures and break brightness-consistency assumption, respectively. To address these problems, in this paper we propose a novel framework with several improvements: (1) we introduce Priors-Based Regularization to learn distribution knowledge from unpaired depth maps and prevent model from being incorrectly trained; (2) we leverage Mapping-Consistent Image Enhancement module to enhance image visibility and contrast while maintaining brightness consistency; and (3) we present Statistics-Based Mask strategy to tune the number of removed pixels within textureless regions, using dynamic statistics. Experimental results demonstrate the effectiveness of each component. Meanwhile, our framework achieves remarkable improvements and state-of-the-art results on two nighttime datasets.

cs.CV↗

arXiv2021

Quantum percolation and magnetic nano-dropletstates in electronically phase-separated manganite nanowires

One-dimensional (1D) confinement has been revealed to effectively tune the properties of materials in homogeneous states. The 1D physics can be further enriched by electronic inhomogeneity, which unfortunately remains largely unknown. Here we demonstrate the ultra-high sensitivity to magnetic fluctuations and the tunability of phase stability in the electronic transport properties of self-assembled electronically phase-separated manganite nanowires with extreme aspect ratio. The onset of magnetic nano-droplet state, a precursor to the ferromagnetic metallic state, is unambiguously revealed, which is attributed to the small lateral size of the nanowires that is comparable to the droplet size. Moreover, the quasi-1D anisotropy stabilizes thin insulating domains to form intrinsic tunneling junctions in the low temperature range, which is robust even under magnetic field up to 14 T, and thus essentially modifies the classic 1D percolation picture to stabilize a novel quantum percolation state. A new phase diagram is therefore established for the manganite system under quasi-1D confinement for the first time. Our findings offer new insight to understand and manipulate the colorful properties of the electronically phase-separated systems via dimensionality engineering.

cond-mat.str-el↗

arXiv2021

Learning to Aggregate and Personalize 3D Face from In-the-Wild Photo Collection

Non-parametric face modeling aims to reconstruct 3D face only from images without shape assumptions. While plausible facial details are predicted, the models tend to over-depend on local color appearance and suffer from ambiguous noise. To address such problem, this paper presents a novel Learning to Aggregate and Personalize (LAP) framework for unsupervised robust 3D face modeling. Instead of using controlled environment, the proposed method implicitly disentangles ID-consistent and scene-specific face from unconstrained photo set. Specifically, to learn ID-consistent face, LAP adaptively aggregates intrinsic face factors of an identity based on a novel curriculum learning approach with relaxed consistency loss. To adapt the face for a personalized scene, we propose a novel attribute-refining network to modify ID-consistent face with target attribute and details. Based on the proposed method, we make unsupervised 3D face modeling benefit from meaningful image facial structure and possibly higher resolutions. Extensive experiments on benchmarks show LAP recovers superior or competitive face shape and texture, compared with state-of-the-art (SOTA) methods with or without prior and supervision.

cs.CV↗

arXiv2021

Efficient Lottery Ticket Finding: Less Data is More

The lottery ticket hypothesis (LTH) reveals the existence of winning tickets (sparse but critical subnetworks) for dense networks, that can be trained in isolation from random initialization to match the latter's accuracies. However, finding winning tickets requires burdensome computations in the train-prune-retrain process, especially on large-scale datasets (e.g., ImageNet), restricting their practical benefits. This paper explores a new perspective on finding lottery tickets more efficiently, by doing so only with a specially selected subset of data, called Pruning-Aware Critical set (PrAC set), rather than using the full training set. The concept of PrAC set was inspired by the recent observation, that deep networks have samples that are either hard to memorize during training, or easy to forget during pruning. A PrAC set is thus hypothesized to capture those most challenging and informative examples for the dense model. We observe that a high-quality winning ticket can be found with training and pruning the dense network on the very compact PrAC set, which can substantially save training iterations for the ticket finding process. Extensive experiments validate our proposal across diverse datasets and network architectures. Specifically, on CIFAR-10, CIFAR-100, and Tiny ImageNet, we locate effective PrAC sets at 35.32%~78.19% of their training set sizes. On top of them, we can obtain the same competitive winning tickets for the corresponding dense networks, yet saving up to 82.85%~92.77%, 63.54%~74.92%, and 76.14%~86.56% training iterations, respectively. Crucially, we show that a PrAC set found is reusable across different network architectures, which can amortize the extra cost of finding PrAC sets, yielding a practical regime for efficient lottery ticket finding.

cs.LG↗

arXiv2021

GANs Can Play Lottery Tickets Too

Deep generative adversarial networks (GANs) have gained growing popularity in numerous scenarios, while usually suffer from high parameter complexities for resource-constrained real-world applications. However, the compression of GANs has less been explored. A few works show that heuristically applying compression techniques normally leads to unsatisfactory results, due to the notorious training instability of GANs. In parallel, the lottery ticket hypothesis shows prevailing success on discriminative models, in locating sparse matching subnetworks capable of training in isolation to full model performance. In this work, we for the first time study the existence of such trainable matching subnetworks in deep GANs. For a range of GANs, we certainly find matching subnetworks at 67%-74% sparsity. We observe that with or without pruning discriminator has a minor effect on the existence and quality of matching subnetworks, while the initialization weights used in the discriminator play a significant role. We then show the powerful transferability of these subnetworks to unseen tasks. Furthermore, extensive experimental results demonstrate that our found subnetworks substantially outperform previous state-of-the-art GAN compression approaches in both image generation (e.g. SNGAN) and image-to-image translation GANs (e.g. CycleGAN). Codes available at https://github.com/VITA-Group/GAN-LTH.

cs.LG↗

arXiv2021

Decentralized Baseband Processing with Gaussian Message Passing Detection for Uplink Massive MU-MIMO Systems

Decentralized baseband processing (DBP) architecture, which partitions the base station antennas into multiple antenna clusters, has been recently proposed to alleviate the excessively high interconnect bandwidth, chip input/output data rates, and detection complexity for massive multi-user multiple-input multiple-output (MU-MIMO) systems. In this paper, we develop a novel decentralized Gaussian message passing (GMP) detection for the DBP architecture. By projecting the discrete probability distribution into a complex Gaussian function, the local means and variances iteratively calculated in each antenna cluster are fused to generate the global symbol beliefs based on the proposed message fusion rule in the centralized processing unit. We present the framework and analysis of the convergence of the decentralized GMP detection based on state evolution under the assumptions of large-system limit and Gaussian sources. Analytical results corroborated by simulations demonstrate that nonuniform antenna cluster partition scheme exhibits higher convergence rate than the uniform counterpart. Simulation results illustrate that the proposed decentralized GMP detection outperforms the recently proposed decentralized algorithms.

eess.SP↗

arXiv2021

Cosmic Velocity Field Reconstruction Using AI

We develop a deep learning technique to infer the non-linear velocity field from the dark matter density field. The deep learning architecture we use is an "U-net" style convolutional neural network, which consists of 15 convolution layers and 2 deconvolution layers. This setup maps the 3-dimensional density field of $32^3$-voxels to the 3-dimensional velocity or momentum fields of $20^3$-voxels. Through the analysis of the dark matter simulation with a resolution of $2 {h^{-1}}{\rm Mpc}$, we find that the network can predict the the non-linearity, complexity and vorticity of the velocity and momentum fields, as well as the power spectra of their value, divergence and vorticity and its prediction accuracy reaches the range of $k\simeq1.4$ $h{\rm Mpc}^{-1}$ with a relative error ranging from 1% to $\lesssim$10%. A simple comparison shows that neural networks may have an overwhelming advantage over perturbation theory in the reconstruction of velocity or momentum fields.

astro-ph.CO↗

arXiv2021

"BNN - BN = ?": Training Binary Neural Networks without Batch Normalization

Batch normalization (BN) is a key facilitator and considered essential for state-of-the-art binary neural networks (BNN). However, the BN layer is costly to calculate and is typically implemented with non-binary parameters, leaving a hurdle for the efficient implementation of BNN training. It also introduces undesirable dependence between samples within each batch. Inspired by the latest advance on Batch Normalization Free (BN-Free) training, we extend their framework to training BNNs, and for the first time demonstrate that BNs can be completed removed from BNN training and inference regimes. By plugging in and customizing techniques including adaptive gradient clipping, scale weight standardization, and specialized bottleneck block, a BN-free BNN is capable of maintaining competitive accuracy compared to its BN-based counterpart. Extensive experiments validate the effectiveness of our proposal across diverse BNN backbones and datasets. For example, after removing BNs from the state-of-the-art ReActNets, it can still be trained with our proposed methodology to achieve 92.08%, 68.34%, and 68.0% accuracy on CIFAR-10, CIFAR-100, and ImageNet respectively, with marginal performance drop (0.23%~0.44% on CIFAR and 1.40% on ImageNet). Codes and pre-trained models are available at: https://github.com/VITA-Group/BNN_NoBN.

cs.LG↗

arXiv2021

Hydrogen-assisted layer-by-layer growth and robust nontrivial topology of stanene films on Bi(111)

Ever since the first successful synthesis of stanene via epitaxial growth, numerous efforts have been devoted to improving its overall quality and exploring its topological and other exotic properties under different growth conditions. Here, using first-principles approaches, we reveal the atomistic growth mechanisms and robust topological properties of few-layer stanene on Bi(111). We first show that monolayer stanene grown on Bi(111) follows a highly desirable nucleation-and-growth mechanism, characterized by attractive interaction of the Sn adatoms. More importantly, we reveal that surface passivation by the residual hydrogen is essential in achieving layer-by-layer growth of high-quality few-layer stanene, with the hydrogen functioning as a surfactant in stabilizing the growing films. Furthermore, we investigate systematically the dependence of the topological properties of stanene on the film thickness, hydrogen passivation, and substrate, and obtain robust quantum spin Hall effects under diverse physical conditions. The robustness of the nontrivial topology is attributable to the strong spin-orbit coupling of the Bi substrate. The atomistic growth mechanisms and nontrivial topology of stanene as presented here are also discussed in connection with recent experimental findings.

cond-mat.mtrl-sci↗

arXiv2021

Electron-Ion Collider in China

Lepton scattering is an established ideal tool for studying inner structure of small particles such as nucleons as well as nuclei. As a future high energy nuclear physics project, an Electron-ion collider in China (EicC) has been proposed. It will be constructed based on an upgraded heavy-ion accelerator, High Intensity heavy-ion Accelerator Facility (HIAF) which is currently under construction, together with a new electron ring. The proposed collider will provide highly polarized electrons (with a polarization of $\sim$80%) and protons (with a polarization of $\sim$70%) with variable center of mass energies from 15 to 20 GeV and the luminosity of (2-3) $\times$ 10$^{33}$ cm$^{-2}$ s$^{-1}$. Polarized deuterons and Helium-3, as well as unpolarized ion beams from Carbon to Uranium, will be also available at the EicC. The main foci of the EicC will be precision measurements of the structure of the nucleon in the sea quark region, including 3D tomography of nucleon; the partonic structure of nuclei and the parton interaction with the nuclear environment; the exotic states, especially those with heavy flavor quark contents. In addition, issues fundamental to understanding the origin of mass could be addressed by measurements of heavy quarkonia near-threshold production at the EicC. In order to achieve the above-mentioned physics goals, a hermetical detector system will be constructed with cutting-edge technologies. This document is the result of collective contributions and valuable inputs from experts across the globe. The EicC physics program complements the ongoing scientific programs at the Jefferson Laboratory and the future EIC project in the United States. The success of this project will also advance both nuclear and particle physics as well as accelerator and detector technology in China.

nucl-ex↗

arXiv2021

Chiral topological superconducting state with Chern number $\mathcal{C} =-2$ in Pb$_3$Bi/Ge(111)

Materials realization of chiral topological superconductivity is a crucial condition for observing and manipulating Majorana fermions in condensed matter physics. Here we develop a tight-binding description of Pb$_3$Bi/Ge(111), identified recently as an appealing candidate system for realizing chiral $p$-wave topological superconductivity [Nat. Phys. 15, 796 (2019)]. We first show that our phenomenological model can capture the two main features of the electronic band structures obtained from first-principles calculations, namely, the giant Rashba splitting and type-II van Hove singularity. Next, when the $s$-wave superconducting property of the parent Pb system is explicitly considered, we find the alloyed system can be tuned into a chiral topological superconductor with Chern number $\mathcal{C} = -2$, resulting from the synergistic effect of a sufficiently strong Zeeman field and the inherently large Rashba spin-orbit coupling. The nontrivial topology with $\mathcal{C} = -2$ is further shown to be detectable as two chiral Majorana edge modes propagating along the same direction of the system with proper boundaries. We finally discuss the physically realistic conditions to establish the predicted topological superconductivity and observe the corresponding Majorana edge modes, including the influence of the superconducting gap, Landé $g$-factor, and critical magnetic field. The present study provides useful guides in searching for effective $p$-wave superconductivity and Majorana fermions in two-dimensional or related interfacial systems.

cond-mat.supr-con↗

arXiv2020

Coarse-to-Fine Pre-training for Named Entity Recognition

More recently, Named Entity Recognition hasachieved great advances aided by pre-trainingapproaches such as BERT. However, currentpre-training techniques focus on building lan-guage modeling objectives to learn a gen-eral representation, ignoring the named entity-related knowledge. To this end, we proposea NER-specific pre-training framework to in-ject coarse-to-fine automatically mined entityknowledge into pre-trained models. Specifi-cally, we first warm-up the model via an en-tity span identification task by training it withWikipedia anchors, which can be deemed asgeneral-typed entities. Then we leverage thegazetteer-based distant supervision strategy totrain the model extract coarse-grained typedentities. Finally, we devise a self-supervisedauxiliary task to mine the fine-grained namedentity knowledge via clustering.Empiricalstudies on three public NER datasets demon-strate that our framework achieves significantimprovements against several pre-trained base-lines, establishing the new state-of-the-art per-formance on three benchmarks. Besides, weshow that our framework gains promising re-sults without using human-labeled trainingdata, demonstrating its effectiveness in label-few and low-resource scenarios

cs.CL↗

arXiv2020

Edge-Enhanced Graph Convolution Networks for Event Detection with Syntactic Relation

Event detection (ED), a key subtask of information extraction, aims to recognize instances of specific event types in text. Previous studies on the task have verified the effectiveness of integrating syntactic dependency into graph convolutional networks. However, these methods usually ignore dependency label information, which conveys rich and useful linguistic knowledge for ED. In this paper, we propose a novel architecture named Edge-Enhanced Graph Convolution Networks (EE-GCN), which simultaneously exploits syntactic structure and typed dependency label information to perform ED. Specifically, an edge-aware node update module is designed to generate expressive word representations by aggregating syntactically-connected words through specific dependency types. Furthermore, to fully explore clues hidden in dependency edges, a node-aware edge update module is introduced, which refines the relation representations with contextual information. These two modules are complementary to each other and work in a mutual promotion way. We conduct experiments on the widely used ACE2005 dataset and the results show significant improvement over competitive baseline methods.

cs.CL↗

arXiv2020

Large-scale inference of correlation among mixed-type biological traits with phylogenetic multivariate probit models

Inferring concerted changes among biological traits along an evolutionary history remains an important yet challenging problem. Besides adjusting for spurious correlation induced from the shared history, the task also requires sufficient flexibility and computational efficiency to incorporate multiple continuous and discrete traits as data size increases. To accomplish this, we jointly model mixed-type traits by assuming latent parameters for binary outcome dimensions at the tips of an unknown tree informed by molecular sequences. This gives rise to a phylogenetic multivariate probit model. With large sample sizes, posterior computation under this model is problematic, as it requires repeated sampling from a high-dimensional truncated normal distribution. Current best practices employ multiple-try rejection sampling that suffers from slow-mixing and a computational cost that scales quadratically in sample size. We develop a new inference approach that exploits 1) the bouncy particle sampler (BPS) based on piecewise deterministic Markov processes to simultaneously sample all truncated normal dimensions, and 2) novel dynamic programming that reduces the cost of likelihood and gradient evaluations for BPS to linear in sample size. In an application with 535 HIV viruses and 24 traits that necessitates sampling from a 12,840-dimensional truncated normal, our method makes it possible to estimate the across-trait correlation and detect factors that affect the pathogen's capacity to cause disease. This inference framework is also applicable to a broader class of covariance structures beyond comparative biology.

stat.ME↗