arXiv subjects

Yang Zhao

Publications and source records attributed to Yang Zhao.

At least 343 records · Page 19Linked to original sources

arXiv2022

Equivalence between algorithmic instability and transition to replica symmetry breaking in perceptron learning systems

Binary perceptron is a fundamental model of supervised learning for the non-convex optimization, which is a root of the popular deep learning. Binary perceptron is able to achieve a classification of random high-dimensional data by computing the marginal probabilities of binary synapses. The relationship between the algorithmic instability and the equilibrium analysis of the model remains elusive. Here, we establish the relationship by showing that the instability condition around the algorithmic fixed point is identical to the instability for breaking the replica symmetric saddle point solution of the free energy function. Therefore, our analysis would hopefully provide insights towards other learning systems in bridging the gap between non-convex learning dynamics and statistical mechanics properties of more complex neural networks.

cond-mat.dis-nn↗

arXiv2022

Neighborhood Region Smoothing Regularization for Finding Flat Minima In Deep Neural Networks

Due to diverse architectures in deep neural networks (DNNs) with severe overparameterization, regularization techniques are critical for finding optimal solutions in the huge hypothesis space. In this paper, we propose an effective regularization technique, called Neighborhood Region Smoothing (NRS). NRS leverages the finding that models would benefit from converging to flat minima, and tries to regularize the neighborhood region in weight space to yield approximate outputs. Specifically, gap between outputs of models in the neighborhood region is gauged by a defined metric based on Kullback-Leibler divergence. This metric provides similar insights with the minimum description length principle on interpreting flat minima. By minimizing both this divergence and empirical loss, NRS could explicitly drive the optimizer towards converging to flat minima. We confirm the effectiveness of NRS by performing image classification tasks across a wide range of model architectures on commonly-used datasets such as CIFAR and ImageNet, where generalization ability could be universally improved. Also, we empirically show that the minima found by NRS would have relatively smaller Hessian eigenvalues compared to the conventional method, which is considered as the evidence of flat minima.

cs.LG↗

arXiv2022

Industrial Experience of Finding Cryptographic Vulnerabilities in Large-scale Codebases

Enterprise environment often screens large-scale (millions of lines of code) codebases with static analysis tools to find bugs and vulnerabilities. Parfait is a static code analysis tool used in Oracle to find security vulnerabilities in industrial codebases. Recently, many studies show that there are complicated cryptographic vulnerabilities caused by misusing cryptographic APIs in Java. In this paper, we describe how we realize a precise and scalable detection of these complicated cryptographic vulnerabilities based on Parfait framework. The key challenge in the detection of cryptographic vulnerabilities is the high false alarm rate caused by pseudo-influences. Pseudo-influences happen if security-irrelevant constants are used in constructing security-critical values. Static analysis is usually unable to distinguish them from hard-coded constants that expose sensitive information. We tackle this problem by specializing the backward dataflow analysis used in Parfait with refinement insights, an idea from the tool CryptoGuard. We evaluate our analyzer on a comprehensive Java cryptographic vulnerability benchmark and eleven large real-world applications. The results show that the Parfait-based cryptographic vulnerability detector can find real-world cryptographic vulnerabilities in large-scale codebases with high true-positive rates and low runtime cost.

cs.SE↗

arXiv2021

SmartDeal: Re-Modeling Deep Network Weights for Efficient Inference and Training

The record-breaking performance of deep neural networks (DNNs) comes with heavy parameterization, leading to external dynamic random-access memory (DRAM) for storage. The prohibitive energy of DRAM accesses makes it non-trivial to deploy DNN on resource-constrained devices, calling for minimizing the weight and data movements to improve the energy efficiency. We present SmartDeal (SD), an algorithm framework to trade higher-cost memory storage/access for lower-cost computation, in order to aggressively boost the storage and energy efficiency, for both inference and training. The core of SD is a novel weight decomposition with structural constraints, carefully crafted to unleash the hardware efficiency potential. Specifically, we decompose each weight tensor as the product of a small basis matrix and a large structurally sparse coefficient matrix whose non-zeros are quantized to power-of-2. The resulting sparse and quantized DNNs enjoy greatly reduced energy for data movement and weight storage, incurring minimal overhead to recover the original weights thanks to the sparse bit-operations and cost-favorable computations. Beyond inference, we take another leap to embrace energy-efficient training, introducing innovative techniques to address the unique roadblocks arising in training while preserving the SD structures. We also design a dedicated hardware accelerator to fully utilize the SD structure to improve the real energy efficiency and latency. We conduct experiments on both multiple tasks, models and datasets in different settings. Results show that: 1) applied to inference, SD achieves up to 2.44x energy efficiency as evaluated via real hardware implementations; 2) applied to training, SD leads to 10.56x and 4.48x reduction in the storage and training energy, with negligible accuracy loss compared to state-of-the-art training baselines. Our source codes are available online.

cs.LG↗

arXiv2021

IRS-Aided SWIPT: Joint Waveform, Active and Passive Beamforming Design Under Nonlinear Harvester Model

The performance of Simultaneous Wireless Information and Power Transfer (SWIPT) is mainly constrained by the received Radio-Frequency (RF) signal strength. To tackle this problem, we introduce an Intelligent Reflecting Surface (IRS) to compensate the propagation loss and boost the transmission efficiency. This paper proposes a novel IRS-aided SWIPT system where a multi-carrier multi-antenna Access Point (AP) transmits information and power simultaneously, with the assist of an IRS, to a single-antenna User Equipment (UE) employing practical receiving schemes. Considering harvester nonlinearity, we characterize the achievable Rate-Energy (R-E) region through a joint optimization of waveform, active and passive beamforming based on the Channel State Information at the Transmitter (CSIT). This problem is solved by the Block Coordinate Descent (BCD) method, where we obtain the active precoder in closed form, the passive beamforming by the Successive Convex Approximation (SCA) approach, and the waveform amplitude by the Geometric Programming (GP) technique. To facilitate practical implementation, we also propose a low-complexity design based on closed-form adaptive waveform schemes. Simulation results demonstrate the proposed algorithms bring considerable R-E gains with robustness to CSIT inaccuracy and finite IRS states, and emphasize the importance of modeling harvester nonlinearity in the IRS-aided SWIPT design.

cs.IT↗

arXiv2021

Performance of a focal plane detector for soft X-ray imaging spectroscopy based on back-illuminated sCMOS

Spectroscopy focusing array (SFA) and Polarimetry focusing array (PFA) are the two major payloads of enhanced X-ray Timing and Polarimetry mission (eXTP). Nested Wolter-\RNum{1} X-ray mirror module is implemented in SFA and PFA to achive high effective area. When evaluating the properties of the mirror module, the alignment of the optical axis of the X-ray mirror module and a quasi-parallel X-ray beam is a prerequisite to ensure the accuracy of the results. Hence, to assist the alignment of the X-ray mirror module, an X-ray focal plane detector is designed based on the back-illuminated scientific Complementary Metal-Oxide-Semiconductor Transistor (sCMOS) sensor GSENSE6060BSI, one of the largest detection areas, is produced by \textit{Gpixel Inc}. Then the characteristics of readout noise, dark current, and split-pixel event properties of the detector are studied with the self-developed multi-target fluorescence X-ray source in a 100 m long X-ray test facility. The energy calibration is carried out with the single-pixel event and the energy non-linearity of the detector is also obtained. Eventually, the simulation of the eXTP mirror module based on the optical model is conducted and the alignment test of the Wolter-\RNum{1} X-ray mirror module designed for \textit{EP/FXT} (Einstein Probe/Follow-up X-ray Telescope) with "Burkert test" method is shown.

physics.ins-det↗

arXiv2021

Large positive magnetoresistance in photocarrier doping potassium tantalites in the extreme quantum limit

We report on a high-field magnetotransport study of KTaO3 single crystals. This material is a promising candidate to study in the extreme quantum limit (EQL). By photocarrier doping with 360 nm light, we have observed a significant positive, non-saturating, and linear magnetoresistance at low temperatures accompanied by a vanishing Hall coefficient. When cooled down to 10 K and subjected to a magnetic field of 12 T, the value of magnetoresistance of KTaO3 (100) is increased by as much as 433%. Such behavior can be attributed to all electrons occupying only the lowest Landau level in the EQL. In this state, a quantum magnetoresistance is produced. This result provides novel insights into the next generation of magnetic devices based on complex materials and adds a new family of materials with positive magnetoresistance.

cond-mat.mtrl-sci↗

arXiv2021

Graph Neural Networks in Real-Time Fraud Detection with Lambda Architecture

Transaction checkout fraud detection is an essential risk control components for E-commerce marketplaces. In order to leverage graph networks to decrease fraud rate efficiently and guarantee the information flow passed through neighbors only from the past of the checkouts, we first present a novel Directed Dynamic Snapshot (DDS) linkage design for graph construction and a Lambda Neural Networks (LNN) architecture for effective inference with Graph Neural Networks embeddings. Experiments show that our LNN on DDS graph, outperforms baseline models significantly and is computational efficient for real-time fraud detection.

cs.LG↗

arXiv2021

Multi-frame Joint Enhancement for Early Interlaced Videos

Early interlaced videos usually contain multiple and interlacing and complex compression artifacts, which significantly reduce the visual quality. Although the high-definition reconstruction technology for early videos has made great progress in recent years, related research on deinterlacing is still lacking. Traditional methods mainly focus on simple interlacing mechanism, and cannot deal with the complex artifacts in real-world early videos. Recent interlaced video reconstruction deep deinterlacing models only focus on single frame, while neglecting important temporal information. Therefore, this paper proposes a multiframe deinterlacing network joint enhancement network for early interlaced videos that consists of three modules, i.e., spatial vertical interpolation module, temporal alignment and fusion module, and final refinement module. The proposed method can effectively remove the complex artifacts in early videos by using temporal redundancy of multi-fields. Experimental results demonstrate that the proposed method can recover high quality results for both synthetic dataset and real-world early interlaced videos.

eess.IV↗

arXiv2021

EAR-NET: Error Attention Refining Network For Retinal Vessel Segmentation

The precise detection of blood vessels in retinal images is crucial to the early diagnosis of the retinal vascular diseases, e.g., diabetic, hypertensive and solar retinopathies. Existing works often fail in predicting the abnormal areas, e.g, sudden brighter and darker areas and are inclined to predict a pixel to background due to the significant class imbalance, leading to high accuracy and specificity while low sensitivity. To that end, we propose a novel error attention refining network (ERA-Net) that is capable of learning and predicting the potential false predictions in a two-stage manner for effective retinal vessel segmentation. The proposed ERA-Net in the refine stage drives the model to focus on and refine the segmentation errors produced in the initial training stage. To achieve this, unlike most previous attention approaches that run in an unsupervised manner, we introduce a novel error attention mechanism which considers the differences between the ground truth and the initial segmentation masks as the ground truth to supervise the attention map learning. Experimental results demonstrate that our method achieves state-of-the-art performance on two common retinal blood vessel datasets.

eess.IV↗

arXiv2021

Rethinking deinterlacing for early interlaced videos

With the rapid development of image restoration techniques, high-definition reconstruction of early videos has achieved impressive results. However, there are few studies about the interlacing artifacts that often appear in early videos and significantly affect visual perception. Traditional deinterlacing approaches are mainly focused on early interlacing scanning systems and thus cannot handle the complex and complicated artifacts in real-world early interlaced videos. Hence, this paper proposes a specific deinterlacing network (DIN), which is motivated by the traditional deinterlacing strategy. The proposed DIN consists of two stages, i.e., a cooperative vertical interpolation stage for split fields, and a merging stage that is applied to perceive movements and remove ghost artifacts. Experimental results demonstrate that the proposed method can effectively remove complex artifacts in early interlaced videos.

cs.CV↗

arXiv2021

RFC-HyPGCN: A Runtime Sparse Feature Compress Accelerator for Skeleton-Based GCNs Action Recognition Model with Hybrid Pruning

Skeleton-based Graph Convolutional Networks (GCNs) models for action recognition have achieved excellent prediction accuracy in the field. However, limited by large model and computation complexity, GCNs for action recognition like 2s-AGCN have insufficient power-efficiency and throughput on GPU. Thus, the demand of model reduction and hardware acceleration for low-power GCNs action recognition application becomes continuously higher. To address challenges above, this paper proposes a runtime sparse feature compress accelerator with hybrid pruning method: RFC-HyPGCN. First, this method skips both graph and spatial convolution workloads by reorganizing the multiplication order. Following spatial convolution workloads channel-pruning dataflow, a coarse-grained pruning method on temporal filters is designed, together with sampling-like fine-grained pruning on time dimension. Later, we come up with an architecture where all convolutional layers are mapped on chip to pursue high throughput. To further reduce storage resource utilization, online sparse feature compress format is put forward. Features are divided and encoded into several banks according to presented format, then bank storage is split into depth-variable mini-banks. Furthermore, this work applies quantization, input-skipping and intra-PE dynamic data scheduling to accelerate the model. In experiments, proposed pruning method is conducted on 2s-AGCN, acquiring 3.0x-8.4x model compression ratio and 73.20\% graph-skipping efficiency with balancing weight pruning. Implemented on Xilinx XCKU-115 FPGA, the proposed architecture has the peak performance of 1142 GOP/s and achieves up to 9.19x and 3.91x speedup over high-end GPU NVIDIA 2080Ti and NVIDIA V100, respectively. Compared with latest accelerator for action recognition GCNs models, our design reaches 22.9x speedup and 28.93\% improvement on DSP efficiency.

cs.AR↗

arXiv2021

Waveform and Beamforming Design for Intelligent Reflecting Surface Aided Wireless Power Transfer: Single-User and Multi-User Solutions

In this paper, we study the waveform and passive beamforming design for intelligent reflecting surface (IRS)-aided wireless power transfer (WPT). Generalized multi-user and low complexity single-user algorithms are derived based on alternating optimization (AO) framework to maximize the weighted sum output DC current, subject to transmit power constraints and passive beamforming phases unit modulus constraints. The input signal waveform and IRS passive beamforming phase shifts are jointly designed as a function of users' individual frequency-selective channel state information (CSI). The energy harvester nonlinearity is explored and two IRS deployment schemes, namely frequency selective IRS (FS-IRS) and frequency flat IRS (FF-IRS), are modeled and analyzed. This paper highlights the fact that IRS can provide an extra passive beamforming gain on output DC power over conventional WPT designs and significantly influence the waveform design by leveraging the benefit of passive beamforming, frequency diversity and energy harvester nonlinearity. Even though FF-IRS exhibits lower output DC current than FS-IRS, it still achieves substantially increased DC power over conventional WPT designs. Performance evaluations confirm the significant benefits of a joint waveform and passive beamforming design accounting for the energy harvester nonlinearity to boost the performance of single-user and multi-user WPT system.

cs.IT↗

arXiv2021

Unpaired Image-to-Image Translation via Latent Energy Transport

Image-to-image translation aims to preserve source contents while translating to discriminative target styles between two visual domains. Most works apply adversarial learning in the ambient image space, which could be computationally expensive and challenging to train. In this paper, we propose to deploy an energy-based model (EBM) in the latent space of a pretrained autoencoder for this task. The pretrained autoencoder serves as both a latent code extractor and an image reconstruction worker. Our model, LETIT, is based on the assumption that two domains share the same latent space, where latent representation is implicitly decomposed as a content code and a domain-specific style code. Instead of explicitly extracting the two codes and applying adaptive instance normalization to combine them, our latent EBM can implicitly learn to transport the source style code to the target style code while preserving the content code, an advantage over existing image translation methods. This simplified solution is also more efficient in the one-sided unpaired image translation setting. Qualitative and quantitative comparisons demonstrate superior translation quality and faithfulness for content preservation. Our model is the first to be applicable to 1024$\times$1024-resolution unpaired image translation to the best of our knowledge.

cs.CV↗

arXiv2021

Phase Diagram of YbZnGaO4 in Applied Magnetic Field

Recently, Yb-based triangular lattice antiferromagnets have garnered significant interest as possible quantum spin liquid candidates. One example is YbMgGaO4, which showed many promising spin liquid features, but also possesses a high degree of disorder owing to site-mixing between the non-magnetic cations. To further elucidate the role of chemical disorder and to explore the phase diagram of these materials in applied field, we present neutron scattering and sensitive magnetometry measurements of the closely related compound, YbZnGaO4. Our results suggest a difference in magnetic anisotropy between the two compounds, and we use key observations of the magnetic phase crossover to motivate an exploration of the field- and exchange parameter-dependent phase diagram, providing an expanded view of the available magnetic states in applied field. This enriched map of the phase space serves as a basis to restrict the values of parameters describing the magnetic Hamiltonian with broad application to recently discovered related materials.

cond-mat.str-el↗

arXiv2021

DNN-Chip Predictor: An Analytical Performance Predictor for DNN Accelerators with Various Dataflows and Hardware Architectures

The recent breakthroughs in deep neural networks (DNNs) have spurred a tremendously increased demand for DNN accelerators. However, designing DNN accelerators is non-trivial as it often takes months/years and requires cross-disciplinary knowledge. To enable fast and effective DNN accelerator development, we propose DNN-Chip Predictor, an analytical performance predictor which can accurately predict DNN accelerators' energy, throughput, and latency prior to their actual implementation. Our Predictor features two highlights: (1) its analytical performance formulation of DNN ASIC/FPGA accelerators facilitates fast design space exploration and optimization; and (2) it supports DNN accelerators with different algorithm-to-hardware mapping methods (i.e., dataflows) and hardware architectures. Experiment results based on 2 DNN models and 3 different ASIC/FPGA implementations show that our DNN-Chip Predictor's predicted performance differs from those of chip measurements of FPGA/ASIC implementation by no more than 17.66% when using different DNN models, hardware architectures, and dataflows. We will release code upon acceptance.

cs.LG↗

arXiv2021

Multi-Scale Context Aggregation Network with Attention-Guided for Crowd Counting

Crowd counting aims to predict the number of people and generate the density map in the image. There are many challenges, including varying head scales, the diversity of crowd distribution across images and cluttered backgrounds. In this paper, we propose a multi-scale context aggregation network (MSCANet) based on single-column encoder-decoder architecture for crowd counting, which consists of an encoder based on a dense context-aware module (DCAM) and a hierarchical attention-guided decoder. To handle the issue of scale variation, we construct the DCAM to aggregate multi-scale contextual information by densely connecting the dilated convolution with varying receptive fields. The proposed DCAM can capture rich contextual information of crowd areas due to its long-range receptive fields and dense scale sampling. Moreover, to suppress the background noise and generate a high-quality density map, we adopt a hierarchical attention-guided mechanism in the decoder. This helps to integrate more useful spatial information from shallow feature maps of the encoder by introducing multiple supervision based on semantic attention module (SAM). Extensive experiments demonstrate that the proposed approach achieves better performance than other similar state-of-the-art methods on three challenging benchmark datasets for crowd counting. The code is available at https://github.com/KingMV/MSCANet

cs.CV↗

arXiv2021

Super-Resolving Compressed Video in Coding Chain

Scaling and lossy coding are widely used in video transmission and storage. Previous methods for enhancing the resolution of such videos often ignore the inherent interference between resolution loss and compression artifacts, which compromises perceptual video quality. To address this problem, we present a mixed-resolution coding framework, which cooperates with a reference-based DCNN. In this novel coding chain, the reference-based DCNN learns the direct mapping from low-resolution (LR) compressed video to their high-resolution (HR) clean version at the decoder side. We further improve reconstruction quality by devising an efficient deformable alignment module with receptive field block to handle various motion distances and introducing a disentangled loss that helps networks distinguish the artifact patterns from texture. Extensive experiments demonstrate the effectiveness of proposed innovations by comparing with state-of-the-art single image, video and reference-based restoration methods.

cs.CV↗