arXiv subjects

Quan Chen

Publications and source records attributed to Quan Chen.

144 records · Page 8Linked to original sources

arXiv2020

How Far Does BERT Look At:Distance-based Clustering and Analysis of BERT$'$s Attention

Recent research on the multi-head attention mechanism, especially that in pre-trained models such as BERT, has shown us heuristics and clues in analyzing various aspects of the mechanism. As most of the research focus on probing tasks or hidden states, previous works have found some primitive patterns of attention head behavior by heuristic analytical methods, but a more systematic analysis specific on the attention patterns still remains primitive. In this work, we clearly cluster the attention heatmaps into significantly different patterns through unsupervised clustering on top of a set of proposed features, which corroborates with previous observations. We further study their corresponding functions through analytical study. In addition, our proposed features can be used to explain and calibrate different attention heads in Transformer models.

cs.CL↗

arXiv2020

Standard $λ$-lattices, rigid ${\rm C}^*$ tensor categories, and (bi)modules

In this article, we construct a 2-shaded rigid ${\rm C}^*$ multitensor category with canonical unitary dual functor directly from a standard $λ$-lattice. We use the notions of traceless Markov towers and lattices to define the notion of module and bimodule over standard $λ$-lattice(s), and we explicitly construct the associated module category and bimodule category over the corresponding 2-shaded rigid ${\rm C}^*$ multitensor category. As an example, we compute the modules and bimodules for Temperley-Lieb-Jones standard $λ$-lattices in terms of traceless Markov towers and lattices. Translating into the unitary 2-category of bigraded Hilbert spaces, we recover DeCommer-Yamshita's classification of $\mathcal{TLJ}$ modules in terms of edge weighted graphs, and a classification of $\mathcal{TLJ}$ bimodules in terms of biunitary connections on square-partite weighted graphs. As an application, we show that every (infinite depth) subfactor planar algebra embeds into the bipartite graph planar algebra of its principal graph.

math.OA↗

arXiv2020

Balancing Efficiency and Flexibility for DNN Acceleration via Temporal GPU-Systolic Array Integration

The research interest in specialized hardware accelerators for deep neural networks (DNN) spikes recently owing to their superior performance and efficiency. However, today's DNN accelerators primarily focus on accelerating specific "kernels" such as convolution and matrix multiplication, which are vital but only part of an end-to-end DNN-enabled application. Meaningful speedups over the entire application often require supporting computations that are, while massively parallel, ill-suited to DNN accelerators. Integrating a general-purpose processor such as a CPU or a GPU incurs significant data movement overhead and leads to resource under-utilization on the DNN accelerators. We propose Simultaneous Multi-mode Architecture (SMA), a novel architecture design and execution model that offers general-purpose programmability on DNN accelerators in order to accelerate end-to-end applications. The key to SMA is the temporal integration of the systolic execution model with the GPU-like SIMD execution model. The SMA exploits the common components shared between the systolic-array accelerator and the GPU, and provides lightweight reconfiguration capability to switch between the two modes in-situ. The SMA achieves up to 63% performance improvement while consuming 23% less energy than the baseline Volta architecture with TensorCore.

cs.DC↗

arXiv2020

Improving Web Content Blocking With Event-Loop-Turn Granularity JavaScript Signatures

Content blocking is an important part of a performant, user-serving, privacy respecting web. Most content blockers build trust labels over URLs. While useful, this approach has well understood shortcomings. Attackers may avoid detection by changing URLs or domains, bundling unwanted code with benign code, or inlining code in pages. The common flaw in existing approaches is that they evaluate code based on its delivery mechanism, not its behavior. In this work we address this problem with a system for generating signatures of the privacy-and-security relevant behavior of executed JavaScript. Our system considers script behavior during each turn on the JavaScript event loop. Focusing on event loop turns allows us to build signatures that are robust against code obfuscation, code bundling, URL modification, and other common evasions, as well as handle unique aspects of web applications. This work makes the following contributions to improving content blocking: First, implement a novel system to build per-event-loop-turn signatures of JavaScript code by instrumenting the Blink and V8 runtimes. Second, we apply these signatures to measure filter list evasion, by using EasyList and EasyPrivacy as ground truth and finding other code that behaves identically. We build ~2m signatures of privacy-and-security behaviors from 11,212 unique scripts blocked by filter lists, and find 3,589 more unique scripts including the same harmful code, affecting 12.48% of websites measured. Third, we taxonomize common filter list evasion techniques. Finally, we present defenses; filter list additions where possible, and a proposed, signature based system in other cases. We share the implementation of our signature-generation system, the dataset from applying our system to the Alexa 100K, and 586 AdBlock Plus compatible filter list rules to block instances of currently blocked code being moved to new URLs.

cs.CR↗

arXiv2020

Towards QoS-Aware and Resource-Efficient GPU Microservices Based on Spatial Multitasking GPUs In Datacenters

While prior researches focus on CPU-based microservices, they are not applicable for GPU-based microservices due to the different contention patterns. It is challenging to optimize the resource utilization while guaranteeing the QoS for GPU microservices. We find that the overhead is caused by inter microservice communication, GPU resource contention and imbalanced throughput within microservice pipeline. We propose Camelot, a runtime system that manages GPU micorservices considering the above factors. In Camelot, a global memory-based communication mechanism enables onsite data sharing that significantly reduces the end-to-end latencies of user queries. We also propose two contention aware resource allocation policies that either maximize the peak supported service load or minimize the resource usage at low load while ensuring the required QoS. The two policies consider the microservice pipeline effect and the runtime GPU resource contention when allocating resources for the microservices. Compared with state-of-the-art work, Camelot increases the supported peak load by up to 64.5% with limited GPUs, and reduces 35% resource usage at low load while achieving the desired 99%-ile latency target.

cs.DC↗

arXiv2020

Topology Virtualization and Dynamics Shielding Method for LEO Satellite Networks

Virtual Node (VN) method is widely adopted to handle satellite network topological dynamics. However, conventional VN method is insufficient when earth rotation and inter-plane phase difference are considered. An improved VN method based on Celestial Sphere Division is proposed to overcome the defects of the conventional method. An optimized inter-satellite link connecting mode is derived to achieve maximal available links. The optimal VN division solution and addressing scheme are designed to generate a nearly static virtual network and solve the asynchronous switches caused by inter-plane phase difference. Comparison results demonstrate the advantages of proposed method.

cs.NI↗

arXiv2019

URSA: Precise Capacity Planning and Contention-aware Scheduling for Public Clouds

Database platform-as-a-service (dbPaaS) is developing rapidly and a large number of databases have been migrated to run on the Clouds for the low cost and flexibility. Emerging Clouds rely on the tenants to provide the resource specification for their database workloads. However, they tend to over-estimate the resource requirement of their databases, resulting in the unnecessarily high cost and low Cloud utilization. A methodology that automatically suggests the "just-enough" resource specification that fulfills the performance requirement of every database workload is profitable. To this end, we propose URSA, a capacity planning and workload scheduling system for dbPaaS Clouds. USRA is comprised by an online capacity planner, a performance interference estimator, and a contention-aware scheduling engine. The capacity planner identifies the most cost-efficient resource specification for a database workload to achieve the required performance online. The interference estimator quantifies the pressure on the shared resource and the sensitivity to the shared resource contention of each database workload. The scheduling engine schedules the workloads across Cloud nodes carefully to eliminate unfair performance interference between the co-located workloads. Our real system experimental results show that URSA reduces 27.5% of CPU usage and 53.4% of memory usage for database workloads while satisfying their performance requirements. Meanwhile, URSA reduces the performance unfairness between the co-located workloads by 42.8% compared with the Kubernetes.

cs.DC↗

arXiv2019

Progressive Feature Polishing Network for Salient Object Detection

Feature matters for salient object detection. Existing methods mainly focus on designing a sophisticated structure to incorporate multi-level features and filter out cluttered features. We present Progressive Feature Polishing Network (PFPN), a simple yet effective framework to progressively polish the multi-level features to be more accurate and representative. By employing multiple Feature Polishing Modules (FPMs) in a recurrent manner, our approach is able to detect salient objects with fine details without any post-processing. A FPM parallelly updates the features of each level by directly incorporating all higher level context information. Moreover, it can keep the dimensions and hierarchical structures of the feature maps, which makes it flexible to be integrated with any CNN-based models. Empirical experiments show that our results are monotonically getting better with increasing number of FPMs. Without bells and whistles, PFPN outperforms the state-of-the-art methods significantly on five benchmark datasets under various evaluation metrics.

cs.CV↗

arXiv2019

Adversarial Defense Through Network Profiling Based Path Extraction

Recently, researchers have started decomposing deep neural network models according to their semantics or functions. Recent work has shown the effectiveness of decomposed functional blocks for defending adversarial attacks, which add small input perturbation to the input image to fool the DNN models. This work proposes a profiling-based method to decompose the DNN models to different functional blocks, which lead to the effective path as a new approach to exploring DNNs' internal organization. Specifically, the per-image effective path can be aggregated to the class-level effective path, through which we observe that adversarial images activate effective path different from normal images. We propose an effective path similarity-based method to detect adversarial images with an interpretable model, which achieve better accuracy and broader applicability than the state-of-the-art technique.

cs.LG↗

arXiv2019

Beta-PtO2: Phononic, thermodynamic, and elastic properties derived from first-principles calculations

Beta-PtO2 is a useful transition metal dioxide, but its fundamental thermodynamic and elastic properties remain unexplored. Using first-principles calculations, we systematically studied the structure, phonon, thermodynamic and elastic properties of beta-PtO2. The lattice dynamics and structural stability of beta-PtO2 under pressure were studied using the phonon spectra and vibrational density of states. The vibrational frequencies of the optical modes of beta-PtO2 increase with elevating pressure; this result is comparable with the available experimental data. Then, the heat capacities and their pressure responses were determined based on the phonon calculations. The pressure dependence of the Debye temperature was studied, and the results are compared in two distinct aspects. The elastic moduli of beta-PtO2 were estimated through the Voigt-Reuss-Hill approximation. The bulk modulus of beta-PtO2 increases linearly with pressure, but the shear modulus is nearly independent of pressure. Our study revealed that the elastic stiffness coefficients C44, C55 and C66 play a primary role in the slow variation of the shear modulus.

cond-mat.mtrl-sci↗

arXiv2019

Prostate Segmentation from 3D MRI Using a Two-Stage Model and Variable-Input Based Uncertainty Measure

This paper proposes a two-stage segmentation model, variable-input based uncertainty measures and an uncertainty-guided post-processing method for prostate segmentation on 3D magnetic resonance images (MRI). The two-stage model was based on 3D dilated U-Nets with the first stage to localize the prostate and the second stage to obtain an accurate segmentation from cropped images. For data augmentation, we proposed the variable-input method which crops the region of interest with additional random variations. Similar to other deep learning models, the proposed model also faced the challenge of suboptimal performance in certain testing cases due to varied training and testing image characteristics. Therefore, it is valuable to evaluate the confidence and performance of the network using uncertainty measures, which are often calculated from the probability maps or their standard deviations with multiple model outputs for the same testing case. However, few studies have quantitatively compared different methods of uncertainty calculation. Furthermore, unlike the commonly used Bayesian dropout during testing, we developed uncertainty measures based on the variable input images at the second stage and evaluated its performance by calculating the correlation with ground-truth-based performance metrics, such as Dice score. For performance estimation, we predicted Dice scores and Hausdorff distance with the most correlated uncertainty measure. For post-processing, we performed Gaussian filter on the underperformed slices to improve segmentation quality. Using PROMISE-12 data, we demonstrated the robustness of the two-stage model and showed high correlation of the proposed variable-input based uncertainty measures with GT-based performance. The uncertainty-guided post-processing method significantly improved label smoothness.

cs.CV↗

arXiv2018

Semantic Human Matting

Human matting, high quality extraction of humans from natural images, is crucial for a wide variety of applications. Since the matting problem is severely under-constrained, most previous methods require user interactions to take user designated trimaps or scribbles as constraints. This user-in-the-loop nature makes them difficult to be applied to large scale data or time-sensitive scenarios. In this paper, instead of using explicit user input constraints, we employ implicit semantic constraints learned from data and propose an automatic human matting algorithm (SHM). SHM is the first algorithm that learns to jointly fit both semantic information and high quality details with deep networks. In practice, simultaneously learning both coarse semantics and fine details is challenging. We propose a novel fusion strategy which naturally gives a probabilistic estimation of the alpha matte. We also construct a very large dataset with high quality annotations consisting of 35,513 unique foregrounds to facilitate the learning and evaluation of human matting. Extensive experiments on this dataset and plenty of real images show that SHM achieves comparable results with state-of-the-art interactive matting methods.

cs.CV↗

arXiv2018

DLFuzz: Differential Fuzzing Testing of Deep Learning Systems

Deep learning (DL) systems are increasingly applied to safety-critical domains such as autonomous driving cars. It is of significant importance to ensure the reliability and robustness of DL systems. Existing testing methodologies always fail to include rare inputs in the testing dataset and exhibit low neuron coverage. In this paper, we propose DLFuzz, the frst differential fuzzing testing framework to guide DL systems exposing incorrect behaviors. DLFuzz keeps minutely mutating the input to maximize the neuron coverage and the prediction difference between the original input and the mutated input, without manual labeling effort or cross-referencing oracles from other DL systems with the same functionality. We present empirical evaluations on two well-known datasets to demonstrate its efficiency. Compared with DeepXplore, the state-of-the-art DL whitebox testing framework, DLFuzz does not require extra efforts to find similar functional DL systems for cross-referencing check, but could generate 338.59% more adversarial inputs with 89.82% smaller perturbations, averagely obtain 2.86% higher neuron coverage, and save 20.11% time consumption.

cs.SE↗

arXiv2018

Deep joint rain and haze removal from single images

Rain removal from a single image is a challenge which has been studied for a long time. In this paper, a novel convolutional neural network based on wavelet and dark channel is proposed. On one hand, we think that rain streaks correspond to high frequency component of the image. Therefore, haar wavelet transform is a good choice to separate the rain streaks and background to some extent. More specifically, the LL subband of a rain image is more inclined to express the background information, while LH, HL, HH subband tend to represent the rain streaks and the edges. On the other hand, the accumulation of rain streaks from long distance makes the rain image look like haze veil. We extract dark channel of rain image as a feature map in network. By increasing this mapping between the dark channel of input and output images, we achieve haze removal in an indirect way. All of the parameters are optimized by back-propagation. Experiments on both synthetic and real- world datasets reveal that our method outperforms other state-of- the-art methods from a qualitative and quantitative perspective.

cs.CV↗

arXiv2017

MSR-net:Low-light Image Enhancement Using Deep Convolutional Network

Images captured in low-light conditions usually suffer from very low contrast, which increases the difficulty of subsequent computer vision tasks in a great extent. In this paper, a low-light image enhancement model based on convolutional neural network and Retinex theory is proposed. Firstly, we show that multi-scale Retinex is equivalent to a feedforward convolutional neural network with different Gaussian convolution kernels. Motivated by this fact, we consider a Convolutional Neural Network(MSR-net) that directly learns an end-to-end mapping between dark and bright images. Different fundamentally from existing approaches, low-light image enhancement in this paper is regarded as a machine learning problem. In this model, most of the parameters are optimized by back-propagation, while the parameters of traditional models depend on the artificial setting. Experiments on a number of challenging images reveal the advantages of our method in comparison with other state-of-the-art methods from the qualitative and quantitative perspective.

cs.CV↗

arXiv2017

Exact 3D seismic data reconstruction using Tubal-Alt-Min algorithm

Data missing is an common issue in seismic data, and many methods have been proposed to solve it. In this paper, we present the low-tubal-rank tensor model and a novel tensor completion algorithm to recover 3D seismic data. This is a fast iterative algorithm, called Tubal-Alt-Min which completes our 3D seismic data by exploiting the low-tubal-rank property expressed as the product of two much smaller tensors. TubalAlt-Min alternates between estimating those two tensor using least squares minimization. We evaluate its reconstruction performance both on synthetic seismic data and land data survey. The experimental results show that compared with the tensor nuclear norm minimization algorithm, Tubal-Alt-Min improves the reconstruction error by orders of magnitude.

cs.CE↗

arXiv2015

EOS: Automatic In-vivo Evolution of Kernel Policies for Better Performance

Today's monolithic kernels often implement a small, fixed set of policies such as disk I/O scheduling policies, while exposing many parameters to let users select a policy or adjust the specific setting of the policy. Ideally, the parameters exposed should be flexible enough for users to tune for good performance, but in practice, users lack domain knowledge of the parameters and are often stuck with bad, default parameter settings. We present EOS, a system that bridges the knowledge gap between kernel developers and users by automatically evolving the policies and parameters in vivo on users' real, production workloads. It provides a simple policy specification API for kernel developers to programmatically describe how the policies and parameters should be tuned, a policy cache to make in-vivo tuning easy and fast by memorizing good parameter settings for past workloads, and a hierarchical search engine to effectively search the parameter space. Evaluation of EOS on four main Linux subsystems shows that it is easy to use and effectively improves each subsystem's performance.

cs.OS↗

arXiv2014

An Approximate Framework for Quantum Transport Calculation with Model Order Reduction

A new approximate computational framework is proposed for computing the non-equilibrium charge density in the context of the non-equilibrium Green's function (NEGF) method for quantum mechanical transport problems. The framework consists of a new formulation, called the X-formulation, for single-energy density calculation based on the solution of sparse linear systems, and a projection-based nonlinear model order reduction (MOR) approach to address the large number of energy points required for large applied biases. The advantages of the new methods are confirmed by numerical experiments.

physics.comp-ph↗