arXiv subjects

Yang Zhao

Publications and source records attributed to Yang Zhao.

At least 379 records · Page 21Linked to original sources

arXiv2020

Dynamic Context Selection for Document-level Neural Machine Translation via Reinforcement Learning

Document-level neural machine translation has yielded attractive improvements. However, majority of existing methods roughly use all context sentences in a fixed scope. They neglect the fact that different source sentences need different sizes of context. To address this problem, we propose an effective approach to select dynamic context so that the document-level translation model can utilize the more useful selected context sentences to produce better translations. Specifically, we introduce a selection module that is independent of the translation module to score each candidate context sentence. Then, we propose two strategies to explicitly select a variable number of context sentences and feed them into the translation module. We train the two modules end-to-end via reinforcement learning. A novel reward is proposed to encourage the selection and utilization of dynamic context sentences. Experiments demonstrate that our approach can select adaptive context sentences for different source sentences, and significantly improves the performance of document-level translation methods.

cs.CL↗

arXiv2020

Q-learning with Language Model for Edit-based Unsupervised Summarization

Unsupervised methods are promising for abstractive text summarization in that the parallel corpora is not required. However, their performance is still far from being satisfied, therefore research on promising solutions is on-going. In this paper, we propose a new approach based on Q-learning with an edit-based summarization. The method combines two key modules to form an Editorial Agent and Language Model converter (EALM). The agent predicts edit actions (e.t., delete, keep, and replace), and then the LM converter deterministically generates a summary on the basis of the action signals. Q-learning is leveraged to train the agent to produce proper edit actions. Experimental results show that EALM delivered competitive performance compared with the previous encoder-decoder-based methods, even with truly zero paired data (i.e., no validation set). Defining the task as Q-learning enables us not only to develop a competitive method but also to make the latest techniques in reinforcement learning available for unsupervised summarization. We also conduct qualitative analysis, providing insights into future study on unsupervised summarizers.

cs.CL↗

arXiv2020

In-plane uniaxial pressure-induced out-of-plane antiferromagnetic moment and critical fluctuations in BaFe$_2$As$_2$

A small in-plane external uniaxial pressure has been widely used as an effective method to acquire single domain iron pnictide BaFe$_2$As$_2$, which exhibits twin-domains without uniaxial strain below the tetragonal-to-orthorhombic structural (nematic) transition temperature $T_s$. Although it is generally assumed that such a pressure will not affect the intrinsic electronic/magnetic properties of the system, it is known to enhance the antiferromagnetic (AF) ordering temperature $T_N$ ($<T_s$) and create in-plane resistivity anisotropy above $T_s$. Here we use neutron polarization analysis to show that such a strain on BaFe$_2$As$_2$ also induces a static or quasi-static out-of-plane ($c$-axis) AF order and its associated critical spin fluctuations near $T_N/T_s$. Therefore, uniaxial pressure necessary to detwin single crystals of BaFe$_2$As$_2$ actually rotates the easy axis of the collinear AF order near $T_N/T_s$, and such effect due to spin-orbit coupling must be taken into account to unveil the intrinsic electronic/magnetic properties of the system.

cond-mat.supr-con↗

arXiv2020

Variational approach to time-dependent fluorescence of a driven qubit

We employ the Dirac-Frenkel variational principle and multiple Davydov ansatz to study time-dependent fluorescence spectra of a driven qubit in the weak- to strong qubit-reservoir coupling regimes, where both the Rabi frequency and spontaneous decay rate are comparable to the transition frequency of the qubit. Our method agrees well with the time-local master-equation approach in the weak-coupling regime, and offers a flexible way to compute the spectra from the bosonic dynamics instead of two-time correlation functions. While the perturbative master equation breaks down in the strong-coupling regime, our method actually becomes more accurate due to the use of bosonic coherent states under certain conditions. We show that the counter-rotating coupling between the qubit and the reservoir has considerable contributions to the photon number dynamics and the spectra under strong driving conditions even though the coupling is moderately weak. The time-dependent spectra are found to be generally asymmetric, a feature that is derived from photon number dynamics. In addition, it is shown that the spectral profiles can be dramatically different from the Mollow triplet due to strong dissipation and/or multiphoton processes associated with the strong driving. Our formalism provides a unique perspective to interpret time-dependent spectra.

quant-ph↗

arXiv2020

Structure-Aware Human-Action Generation

Generating long-range skeleton-based human actions has been a challenging problem since small deviations of one frame can cause a malformed action sequence. Most existing methods borrow ideas from video generation, which naively treat skeleton nodes/joints as pixels of images without considering the rich inter-frame and intra-frame structure information, leading to potential distorted actions. Graph convolutional networks (GCNs) is a promising way to leverage structure information to learn structure representations. However, directly adopting GCNs to tackle such continuous action sequences both in spatial and temporal spaces is challenging as the action graph could be huge. To overcome this issue, we propose a variant of GCNs to leverage the powerful self-attention mechanism to adaptively sparsify a complete action graph in the temporal space. Our method could dynamically attend to important past frames and construct a sparse graph to apply in the GCN framework, well-capturing the structure information in action sequences. Extensive experimental results demonstrate the superiority of our method on two standard human action datasets compared with existing methods.

cs.GR↗

arXiv2020

Magnetic Order and Competition With Superconductivity in (Er-Ho)Ni$_{2}$B$_{2}$C

The rare earth magnetic order in pure and doped Er$_{(1-x)}$Ho$_{x}$Ni$_2$B$_2$C (x~=~0,~0.25,~0.50,~0.75,~1) single crystal samples was investigated using magnetization and neutron diffraction measurements. Superconducting quaternary borocarbides, $R$Ni$_2$B$_2$C where R~=~Er, Ho , are both magnetic intermetallic superconductors with the transition temperatures $\sim$ 10 K. These compounds also develop magnetic order in the vicinity of this temperature. Depending on the rare earth composition the coupling between superconductivity and magnetism creates several phases, ranging from a reentrant superconductor with a mixture of commensurate and incommensurate antiferromagnetism to a total incommensurate antiferromagnetic spin modulation with a weak ferromagnetic state. All of these phases coexist with superconductivity. RKKY magnetic interactions are used to describe the magnetic orders in the pure compounds. However, the doping of Ho on Er (or Er on Ho) sites which have two strong magnetic moments with two different easy directions creates new and complicated magnetic modulations with possible local disorder effects. One fascinating effect is the development of an induced magnetic state resembling the pure and doped $R_2$CuO$_4$, R~=~Nd and Pr.

cond-mat.supr-con↗

arXiv2020

Attacks to Federated Learning: Responsive Web User Interface to Recover Training Data from User Gradients

Local differential privacy (LDP) is an emerging privacy standard to protect individual user data. One scenario where LDP can be applied is federated learning, where each user sends in his/her user gradients to an aggregator who uses these gradients to perform stochastic gradient descent. In a case where the aggregator is untrusted and LDP is not applied to each user gradient, the aggregator can recover sensitive user data from these gradients. In this paper, we present a new interactive web demo showcasing the power of local differential privacy by visualizing federated learning with local differential privacy. Moreover, the live demo shows how LDP can prevent untrusted aggregators from recovering sensitive training data. A measure called the exp-hamming recovery is also created to show the extent of how much data the aggregator can recover.

cs.CR↗

arXiv2020

Feature Quantization Improves GAN Training

The instability in GAN training has been a long-standing problem despite remarkable research efforts. We identify that instability issues stem from difficulties of performing feature matching with mini-batch statistics, due to a fragile balance between the fixed target distribution and the progressively generated distribution. In this work, we propose Feature Quantization (FQ) for the discriminator, to embed both true and fake data samples into a shared discrete space. The quantized values of FQ are constructed as an evolving dictionary, which is consistent with feature statistics of the recent distribution history. Hence, FQ implicitly enables robust feature matching in a compact space. Our method can be easily plugged into existing GAN models, with little computational overhead in training. We apply FQ to 3 representative GAN models on 9 benchmarks: BigGAN for image generation, StyleGAN for face synthesis, and U-GAT-IT for unsupervised image-to-image translation. Extensive experimental results show that the proposed FQ-GAN can improve the FID scores of baseline methods by a large margin on a variety of tasks, achieving new state-of-the-art performance.

cs.LG↗

arXiv2020

DeGNN: Characterizing and Improving Graph Neural Networks with Graph Decomposition

Despite the wide application of Graph Convolutional Network (GCN), one major limitation is that it does not benefit from the increasing depth and suffers from the oversmoothing problem. In this work, we first characterize this phenomenon from the information-theoretic perspective and show that under certain conditions, the mutual information between the output after $l$ layers and the input of GCN converges to 0 exponentially with respect to $l$. We also show that, on the other hand, graph decomposition can potentially weaken the condition of such convergence rate, which enabled our analysis for GraphCNN. While different graph structures can only benefit from the corresponding decomposition, in practice, we propose an automatic connectivity-aware graph decomposition algorithm, DeGNN, to improve the performance of general graph neural networks. Extensive experiments on widely adopted benchmark datasets demonstrate that DeGNN can not only significantly boost the performance of corresponding GNNs, but also achieves the state-of-the-art performances.

cs.LG↗

arXiv2020

AutoDNNchip: An Automated DNN Chip Predictor and Builder for Both FPGAs and ASICs

Recent breakthroughs in Deep Neural Networks (DNNs) have fueled a growing demand for DNN chips. However, designing DNN chips is non-trivial because: (1) mainstream DNNs have millions of parameters and operations; (2) the large design space due to the numerous design choices of dataflows, processing elements, memory hierarchy, etc.; and (3) an algorithm/hardware co-design is needed to allow the same DNN functionality to have a different decomposition, which would require different hardware IPs to meet the application specifications. Therefore, DNN chips take a long time to design and require cross-disciplinary experts. To enable fast and effective DNN chip design, we propose AutoDNNchip - a DNN chip generator that can automatically generate both FPGA- and ASIC-based DNN chip implementation given DNNs from machine learning frameworks (e.g., PyTorch) for a designated application and dataset. Specifically, AutoDNNchip consists of two integrated enablers: (1) a Chip Predictor, built on top of a graph-based accelerator representation, which can accurately and efficiently predict a DNN accelerator's energy, throughput, and area based on the DNN model parameters, hardware configuration, technology-based IPs, and platform constraints; and (2) a Chip Builder, which can automatically explore the design space of DNN chips (including IP selection, block configuration, resource balancing, etc.), optimize chip design via the Chip Predictor, and then generate optimized synthesizable RTL to achieve the target design metrics. Experimental results show that our Chip Predictor's predicted performance differs from real-measured ones by < 10% when validated using 15 DNN models and 4 platforms (edge-FPGA/TPU/GPU and ASIC). Furthermore, accelerators generated by our AutoDNNchip can achieve better (up to 3.86X improvement) performance than that of expert-crafted state-of-the-art accelerators.

cs.DC↗

arXiv2020

Blockchain-Based Differential Privacy Cost Management System

Privacy preservation is a big concern for various sectors. To protect individual user data, one emerging technology is differential privacy. However, it still has limitations for datasets with frequent queries, such as the fast accumulation of privacy cost. To tackle this limitation, this paper explores the integration of a secured decentralised ledger, blockchain. Blockchain will be able to keep track of all noisy responses generated with differential privacy algorithm and allow for certain queries to reuse old responses. In this paper, a demo of a proposed blockchain-based privacy management system is designed as an interactive decentralised web application (DApp). The demo created illustrates that leveraging on blockchain will allow the total privacy cost accumulated to decrease significantly.

cs.CR↗

arXiv2020

Correlating magnetic structure and magnetotransport in semimetal thin films of Eu$_{1-x}$Sm$_x$TiO$_3$

We report on the evolution of the average and depth-dependent magnetic order in thin film samples of biaxially stressed and electron-doped EuTiO$_3$ for samples across a doping range $<$0.1 to 7.8 $\times 10^{20}$ cm$^{-3}$. Under an applied in-plane magnetic field, the G-type antiferromagnetic ground state undergoes a continuous spin-flop phase transition into in-plane, field-polarized ferromagnetism. The critical field for ferromagnetism slightly decreases with an increasing number of free carriers, yet the field evolution of the spin-flop transition is qualitatively similar across the doping range. Unexpectedly, we observe interfacial ferromagnetism with saturated Eu$^{2+}$ moments at the substrate interface at low fields preceding ferromagnetic saturation throughout the bulk of the degenerate semiconductor film. We discuss the implications of these findings for the unusual magnetotransport properties of this compound.

cond-mat.str-el↗

arXiv2020

Topological Singularity Induced Chiral Kohn Anomaly in a Weyl Semimetal

The electron-phonon interaction (EPI) is instrumental in a wide variety of phenomena in solid-state physics, such as electrical resistivity in metals, carrier mobility, optical transition and polaron effects in semiconductors, lifetime of hot carriers, transition temperature in BCS superconductors, and even spin relaxation in diamond nitrogen-vacancy centers for quantum information processing. However, due to the weak EPI strength, most phenomena have focused on electronic properties rather than on phonon properties. One prominent exception is the Kohn anomaly, where phonon softening can emerge when the phonon wavevector nests the Fermi surface of metals. Here we report a new class of Kohn anomaly in a topological Weyl semimetal (WSM), predicted by field-theoretical calculations, and experimentally observed through inelastic x-ray and neutron scattering on WSM tantalum phosphide (TaP). Compared to the conventional Kohn anomaly, the Fermi surface in a WSM exhibits multiple topological singularities of Weyl nodes, leading to a distinct nesting condition with chiral selection, a power-law divergence, and non-negligible dynamical effects. Our work brings the concept of Kohn anomaly into WSMs and sheds light on elucidating the EPI mechanism in emergent topological materials.

cond-mat.mtrl-sci↗

arXiv2020

A New MRAM-based Process In-Memory Accelerator for Efficient Neural Network Training with Floating Point Precision

The excellent performance of modern deep neural networks (DNNs) comes at an often prohibitive training cost, limiting the rapid development of DNN innovations and raising various environmental concerns. To reduce the dominant data movement cost of training, process in-memory (PIM) has emerged as a promising solution as it alleviates the need to access DNN weights. However, state-of-the-art PIM DNN training accelerators employ either analog/mixed signal computing which has limited precision or digital computing based on a memory technology that supports limited logic functions and thus requires complicated procedure to realize floating point computation. In this paper, we propose a spin orbit torque magnetic random access memory (SOT-MRAM) based digital PIM accelerator that supports floating point precision. Specifically, this new accelerator features an innovative (1) SOT-MRAM cell, (2) full addition design, and (3) floating point computation. Experiment results show that the proposed SOT-MRAM PIM based DNN training accelerator can achieve 3.3$\times$, 1.8$\times$, and 2.5$\times$ improvement in terms of energy, latency, and area, respectively, compared with a state-of-the-art PIM based DNN training accelerator.

cs.DC↗

arXiv2020

SmartExchange: Trading Higher-cost Memory Storage/Access for Lower-cost Computation

We present SmartExchange, an algorithm-hardware co-design framework to trade higher-cost memory storage/access for lower-cost computation, for energy-efficient inference of deep neural networks (DNNs). We develop a novel algorithm to enforce a specially favorable DNN weight structure, where each layerwise weight matrix can be stored as the product of a small basis matrix and a large sparse coefficient matrix whose non-zero elements are all power-of-2. To our best knowledge, this algorithm is the first formulation that integrates three mainstream model compression ideas: sparsification or pruning, decomposition, and quantization, into one unified framework. The resulting sparse and readily-quantized DNN thus enjoys greatly reduced energy consumption in data movement as well as weight storage. On top of that, we further design a dedicated accelerator to fully utilize the SmartExchange-enforced weights to improve both energy efficiency and latency performance. Extensive experiments show that 1) on the algorithm level, SmartExchange outperforms state-of-the-art compression techniques, including merely sparsification or pruning, decomposition, and quantization, in various ablation studies based on nine DNN models and four datasets; and 2) on the hardware level, the proposed SmartExchange based accelerator can improve the energy efficiency by up to 6.7$\times$ and the speedup by up to 19.2$\times$ over four state-of-the-art DNN accelerators, when benchmarked on seven DNN models (including four standard DNNs, two compact DNN models, and one segmentation model) and three datasets.

cs.LG↗

arXiv2020

TIMELY: Pushing Data Movements and Interfaces in PIM Accelerators Towards Local and in Time Domain

Resistive-random-access-memory (ReRAM) based processing-in-memory (R$^2$PIM) accelerators show promise in bridging the gap between Internet of Thing devices' constrained resources and Convolutional/Deep Neural Networks' (CNNs/DNNs') prohibitive energy cost. Specifically, R$^2$PIM accelerators enhance energy efficiency by eliminating the cost of weight movements and improving the computational density through ReRAM's high density. However, the energy efficiency is still limited by the dominant energy cost of input and partial sum (Psum) movements and the cost of digital-to-analog (D/A) and analog-to-digital (A/D) interfaces. In this work, we identify three energy-saving opportunities in R$^2$PIM accelerators: analog data locality, time-domain interfacing, and input access reduction, and propose an innovative R$^2$PIM accelerator called TIMELY, with three key contributions: (1) TIMELY adopts analog local buffers (ALBs) within ReRAM crossbars to greatly enhance the data locality, minimizing the energy overheads of both input and Psum movements; (2) TIMELY largely reduces the energy of each single D/A (and A/D) conversion and the total number of conversions by using time-domain interfaces (TDIs) and the employed ALBs, respectively; (3) we develop an only-once input read (O$^2$IR) mapping method to further decrease the energy of input accesses and the number of D/A conversions. The evaluation with more than 10 CNN/DNN models and various chip configurations shows that, TIMELY outperforms the baseline R$^2$PIM accelerator, PRIME, by one order of magnitude in energy efficiency while maintaining better computational density (up to 31.2$\times$) and throughput (up to 736.6$\times$). Furthermore, comprehensive studies are performed to evaluate the effectiveness of the proposed ALB, TDI, and O$^2$IR innovations in terms of energy savings and area reduction.

cs.LG↗

arXiv2020

3D Face Reconstruction from A Single Image Assisted by 2D Face Images in the Wild

3D face reconstruction from a single 2D image is a challenging problem with broad applications. Recent methods typically aim to learn a CNN-based 3D face model that regresses coefficients of 3D Morphable Model (3DMM) from 2D images to render 3D face reconstruction or dense face alignment. However, the shortage of training data with 3D annotations considerably limits performance of those methods. To alleviate this issue, we propose a novel 2D-assisted self-supervised learning (2DASL) method that can effectively use "in-the-wild" 2D face images with noisy landmark information to substantially improve 3D face model learning. Specifically, taking the sparse 2D facial landmarks as additional information, 2DSAL introduces four novel self-supervision schemes that view the 2D landmark and 3D landmark prediction as a self-mapping process, including the 2D and 3D landmark self-prediction consistency, cycle-consistency over the 2D landmark prediction and self-critic over the predicted 3DMM coefficients based on landmark predictions. Using these four self-supervision schemes, the 2DASL method significantly relieves demands on the the conventional paired 2D-to-3D annotations and gives much higher-quality 3D face models without requiring any additional 3D annotations. Experiments on multiple challenging datasets show that our method outperforms state-of-the-arts for both 3D face reconstruction and dense face alignment by a large margin.

cs.CV↗

arXiv2020

Materializing Rival Ground States in the Barlowite Family of Kagome Magnets: Quantum Spin Liquid, Spin Ordered, and Valence Bond Crystal States

The spin-$\frac{1}{2}$ kagome antiferromagnet is considered an ideal host for a quantum spin liquid ground state. We find that when the bonds of the kagome lattice are modulated with a periodic pattern, new quantum ground states emerge. Newly synthesized crystalline barlowite (Cu$_4$(OH)$_6$FBr) and Zn-substituted barlowite demonstrate the delicate interplay between singlet states and spin order on the spin-$\frac{1}{2}$ kagome lattice. Comprehensive structural measurements demonstrate that our new variant of barlowite maintains hexagonal symmetry at low temperatures with an arrangement of distorted and undistorted kagome triangles, for which numerical simulations predict a pinwheel valence bond crystal (VBC) state instead of a quantum spin liquid (QSL). The presence of interlayer spins eventually leads to an interesting pinwheel $q=0$ magnetic order. Partially Zn-substituted barlowite (Cu$_{3.44}$Zn$_{0.56}$(OH)$_6$FBr) has an ideal kagome lattice and shows QSL behavior, indicating a surprising robustness of the QSL against interlayer impurities. The magnetic susceptibility is similar to that of herbertsmithite, even though the Cu$^{2+}$ impurities are above the percolation threshold for the interlayer lattice and they couple more strongly to the nearest kagome moment. This system is a unique playground displaying QSL, VBC, and spin order, furthering our understanding of these highly competitive quantum states.

cond-mat.str-el↗