arXiv subjects

He Zhang

Publications and source records attributed to He Zhang.

At least 253 records · Page 14Linked to original sources

arXiv2021

Exploiting the Unique Expression for Improved Sentiment Analysis in Software Engineering Text

Sentiment analysis on software engineering (SE) texts has been widely used in the SE research, such as evaluating app reviews or analyzing developers sentiments in commit messages. To better support the use of automated sentiment analysis for SE tasks, researchers built an SE-domain-specified sentiment dictionary to further improve the accuracy of the results. Unfortunately, recent work reported that current mainstream tools for sentiment analysis still cannot provide reliable results when analyzing the sentiments in SE texts. We suggest that the reason for this situation is because the way of expressing sentiments in SE texts is largely different from the way in social network or movie comments. In this paper, we propose to improve sentiment analysis in SE texts by using sentence structures, a different perspective from building a domain dictionary. Specifically, we use sentence structures to first identify whether the author is expressing her sentiment in a given clause of an SE text, and to further adjust the calculation of sentiments which are confirmed in the clause. An empirical evaluation based on four different datasets shows that our approach can outperform two dictionary-based baseline approaches, and is more generalizable compared to a learning-based baseline approach.

cs.SE↗

arXiv2021

CrowdSim: A Hybrid Simulation Model for Failure Prediction in Crowdsourced Software Development

A typical crowdsourcing software development(CSD) marketplace consists of a list of software tasks as service demands and a pool of freelancer developers as service suppliers. Highly dynamic and competitive CSD market places may result in task failure due to unforeseen risks, such as increased competition over shared worker supply, or uncertainty associated with workers' experience and skills, and so on. To improve CSD effectiveness, it is essential to better understand and plan with respect to dynamic worker characteristics and risks associated with CSD processes. In this paper, we present a hybrid simulation model, CrowdSim, to forecast crowdsourcing task failure risk in competitive CSD platforms. CrowdSim is composed of three layered components: the macro-level reflects the overall crowdsourcing platform based on system dynamics,the meso-level represents the task life cycle based on discrete event simulation, and the micro-level models the crowd workers' decision-making processes based on agent-based simulation. CrowdSim is evaluated through three CSD decision scenarios to demonstrate its effectiveness, using a real-world historical dataset and the results demonstrate CrowdSim's potential in empowering crowdsourcing managers to explore crowdsourcing outcomes with respect to different task scheduling options.

cs.SE↗

arXiv2021

Solving Backward Doubly Stochastic Differential Equations through Splitting Schemes

A splitting scheme for backward doubly stochastic differential equations is proposed. The main idea is to decompose a backward doubly stochastic differential equation into a backward stochastic differential equation and a stochastic differential equation. The backward stochastic differential equation and the stochastic differential equation are then approximated by first order finite difference schemes, which results in a first order scheme for the backward doubly stochastic differential equation. Numerical experiments are conducted to illustrate the convergence rate of the proposed scheme.

math.NA↗

arXiv2021

Using Relative Lines of Code to Guide Automated Test Generation for Python

Raw lines of code (LOC) is a metric that does not, at first glance, seem extremely useful for automated test generation. It is both highly language-dependent and not extremely meaningful, semantically, within a language: one coder can produce the same effect with many fewer lines than another. However, relative LOC, between components of the same project, turns out to be a highly useful metric for automated testing. In this paper, we make use of a heuristic based on LOC counts for tested functions to dramatically improve the effectiveness of automated test generation. This approach is particularly valuable in languages where collecting code coverage data to guide testing has a very high overhead.We apply the heuristic to property-based Python testing using the TSTL (Template Scripting Testing Language) tool. In our experiments, the simple LOC heuristic can improve branch and statement coverage by large margins (often more than 20%, up to 40% or more), and improve fault detection by an even larger margin (usually more than 75%, and up to 400% or more). The LOC heuristic is also easy to combine with other approaches, and is comparable to, and possibly more effective than, two well-established approaches for guiding random testing.

cs.SE↗

arXiv2021

A Driving Behavior Recognition Model with Bi-LSTM and Multi-Scale CNN

In autonomous driving, perceiving the driving behaviors of surrounding agents is important for the ego-vehicle to make a reasonable decision. In this paper, we propose a neural network model based on trajectories information for driving behavior recognition. Unlike existing trajectory-based methods that recognize the driving behavior using the hand-crafted features or directly encoding the trajectory, our model involves a Multi-Scale Convolutional Neural Network (MSCNN) module to automatically extract the high-level features which are supposed to encode the rich spatial and temporal information. Given a trajectory sequence of an agent as the input, firstly, the Bi-directional Long Short Term Memory (Bi-LSTM) module and the MSCNN module respectively process the input, generating two features, and then the two features are fused to classify the behavior of the agent. We evaluate the proposed model on the public BLVD dataset, achieving a satisfying performance.

cs.CV↗

arXiv2021

Robust and Efficient Single-Pixel Image Classificationwith Nonlinear Optics

We present a hybrid image classifier by mode-selective image upconversion, single pixel photodetection, and deep learning, aiming at fast processing a large number of pixels. It utilizes partial Fourier transform to extract the signature features of images in both the original and Fourier domains, thereby significantly increasing the classification accuracy and robustness. Tested on the MNIST handwritten digit images, it boosts the accuracy from 81.25% to 99.23%, and achieves an 83% accuracy for highly contaminated images whose signal-to-noise ratio is only -17 dB. Our approach could prove useful for fast lidar data processing, high resolution image recognition, occluded target identification, atmosphere monitoring, and so on.

physics.optics↗

arXiv2020

Meticulous Object Segmentation

Compared with common image segmentation tasks targeted at low-resolution images, higher resolution detailed image segmentation receives much less attention. In this paper, we propose and study a task named Meticulous Object Segmentation (MOS), which is focused on segmenting well-defined foreground objects with elaborate shapes in high resolution images (e.g. 2k - 4k). To this end, we propose the MeticulousNet which leverages a dedicated decoder to capture the object boundary details. Specifically, we design a Hierarchical Point-wise Refining (HierPR) block to better delineate object boundaries, and reformulate the decoding process as a recursive coarse to fine refinement of the object mask. To evaluate segmentation quality near object boundaries, we propose the Meticulosity Quality (MQ) score considering both the mask coverage and boundary precision. In addition, we collect a MOS benchmark dataset including 600 high quality images with complex objects. We provide comprehensive empirical evidence showing that MeticulousNet can reveal pixel-accurate segmentation boundaries and is superior to state-of-the-art methods for high resolution object segmentation tasks.

cs.CV↗

arXiv2020

Estimating linear response statistics using orthogonal polynomials: An RKHS formulation

We study the problem of estimating linear response statistics under external perturbations using time series of unperturbed dynamics. Based on the fluctuation-dissipation theory, this problem is reformulated as an unsupervised learning task of estimating a density function. We consider a nonparametric density estimator formulated by the kernel embedding of distributions with "Mercer-type" kernels, constructed based on the classical orthogonal polynomials defined on non-compact domains. While the resulting representation is analogous to Polynomial Chaos Expansion (PCE), the connection to the reproducing kernel Hilbert space (RKHS) theory allows one to establish the uniform convergence of the estimator and to systematically address a practical question of identifying the PCE basis for a consistent estimation. We also provide practical conditions for the well-posedness of not only the estimator but also of the underlying response statistics. Finally, we provide a statistical error bound for the density estimation that accounts for the Monte-Carlo averaging over non-i.i.d time series and the biases due to a finite basis truncation. This error bound provides a means to understand the feasibility as well as limitation of the kernel embedding with Mercer-type kernels. Numerically, we verify the effectiveness of the estimator on two stochastic dynamics with known, yet, non-trivial equilibrium densities.

math.ST↗

arXiv2020

Linear Response Based Parameter Estimation in the Presence of Model Error

Recently, we proposed a method to estimate parameters of stochastic dynamics based on the linear response statistics. The method rests upon a nonlinear least-squares problem that takes into account the response properties that stem from the Fluctuation-Dissipation Theory. In this article, we address an important issue that arises in the presence of model error. In particular, when the equilibrium density function is high dimensional and non-Gaussian, and in some cases, is unknown, the linear response statistics are inaccessible. We show that this issue can be resolved by fitting the imperfect model to appropriate marginal linear response statistics that can be approximated using the available data and parametric or nonparametric models. The effectiveness of the parameter estimation approach is demonstrated in the context of molecular dynamical models (Langevin dynamics) with a non-uniform temperature profile, where the modeling error is due to coarse-graining, and a PDE (non-Langevin dynamics) that exhibits spatiotemporal chaos, where the model error arises from a severe spectral truncation. In these examples, we show how the imperfect models, the Langevin equation with parameters estimated using the proposed scheme, can predict the nonlinear response statistics of the underlying dynamics under admissible external disturbances.

math.NA↗

arXiv2020

Deep Image Compositing

Image compositing is a task of combining regions from different images to compose a new image. A common use case is background replacement of portrait images. To obtain high quality composites, professionals typically manually perform multiple editing steps such as segmentation, matting and foreground color decontamination, which is very time consuming even with sophisticated photo editing tools. In this paper, we propose a new method which can automatically generate high-quality image compositing without any user input. Our method can be trained end-to-end to optimize exploitation of contextual and color information of both foreground and background images, where the compositing quality is considered in the optimization. Specifically, inspired by Laplacian pyramid blending, a dense-connected multi-stream fusion network is proposed to effectively fuse the information from the foreground and background images at different scales. In addition, we introduce a self-taught strategy to progressively train from easy to complex cases to mitigate the lack of training data. Experiments show that the proposed method can automatically generate high-quality composites and outperforms existing methods both qualitatively and quantitatively.

cs.CV↗

arXiv2020

Single-Pixel Pattern Recognition with Coherent Nonlinear Optics

We propose and experimentally demonstrate a nonlinear-optics approach to pattern recognition with single-pixel imaging and deep neural network. It employs mode selective image up-conversion to project a raw image onto a set of coherent spatial modes, whereby its signature features are extracted nonlinear-optically. With 40 projection modes, the classification accuracy reaches a high value of 99.49% for the MNIST handwritten digit images, and up to 95.32% even when they are mixed with strong noise. Our experiment harnesses rich coherent processes in nonlinear optics for efficient machine learning, with potential applications in online classification of large size images, fast lidar data analyses, complex pattern recognition, and so on.

physics.optics↗

arXiv2020

Synergy between Machine/Deep Learning and Software Engineering: How Far Are We?

Since 2009, the deep learning revolution, which was triggered by the introduction of ImageNet, has stimulated the synergy between Machine Learning (ML)/Deep Learning (DL) and Software Engineering (SE). Meanwhile, critical reviews have emerged that suggest that ML/DL should be used cautiously. To improve the quality (especially the applicability and generalizability) of ML/DL-related SE studies, and to stimulate and enhance future collaborations between SE/AI researchers and industry practitioners, we conducted a 10-year Systematic Literature Review (SLR) on 906 ML/DL-related SE papers published between 2009 and 2018. Our trend analysis demonstrated the mutual impacts that ML/DL and SE have had on each other. At the same time, however, we also observed a paucity of replicable and reproducible ML/DL-related SE studies and identified five factors that influence their replicability and reproducibility. To improve the applicability and generalizability of research results, we analyzed what ingredients in a study would facilitate an understanding of why a ML/DL technique was selected for a specific SE problem. In addition, we identified the unique trends of impacts of DL models on SE tasks, as well as five unique challenges that needed to be met in order to better leverage DL to improve the productivity of SE tasks. Finally, we outlined a road-map that we believe can facilitate the transfer of ML/DL-based SE research results into real-world industry practices.

cs.SE↗

arXiv2020

Efficient mesh deformation using radial basis functions with a grouping-circular-based greedy algorithm

A grouping-circular-based (GCB) greedy algorithm is proposed to promote the efficiency of mesh deformation. By incorporating the multigrid concept that the computational errors on the fine mesh can be approximated with those on the coarse mesh, this algorithm stochastically divides all boundary nodes into $m$ groups and uses the locally maximum radial basis functions (RBF) interpolation error of each group as an approximation to the globally maximum one of all boundary nodes in each iterative procedure for reducing the RBF support nodes. For this reason, it avoids the interpolation conducted at all boundary nodes and thus reduces the corresponding computational complexity from $O\left({N_c^2{N_b}} \right)$ to $O\left( {N_c^3} \right)$. Besides, after $m$ iterations, the interpolation errors of all boundary nodes are computed once, thus allowing all boundary nodes can contribute to error control. Two canonical deformation problems of the ONERA M6 wing and the DLR-F6 Wing-Body-Nacelle-Pylon configuration are computed to validate the GCB greedy algorithm. The computational results show that the GCB greedy algorithm is able to remarkably promote the efficiency of computing the interpolation errors in the data reducing procedure by dozens of times. Because an increase of $m$ results in an increase of $N_c$, an appropriate range of $\left[ {{N_b}/{N_c},{\rm{ }}2{N_b}/{N_c}}\right]$ for $m$ is suggested to prevent too much additional computations for solving the linear algebraic system and computing the displacements of volume nodes induced by the increase of $N_c $. The results also show that the GCB greedy algorithm tends to generate a more significant efficiency improvement for mesh deformation when a larger-scale mesh is applied.

math.NA↗

arXiv2020

SummPip: Unsupervised Multi-Document Summarization with Sentence Graph Compression

Obtaining training data for multi-document summarization (MDS) is time consuming and resource-intensive, so recent neural models can only be trained for limited domains. In this paper, we propose SummPip: an unsupervised method for multi-document summarization, in which we convert the original documents to a sentence graph, taking both linguistic and deep representation into account, then apply spectral clustering to obtain multiple clusters of sentences, and finally compress each cluster to generate the final summary. Experiments on Multi-News and DUC-2004 datasets show that our method is competitive to previous unsupervised methods and is even comparable to the neural supervised approaches. In addition, human evaluation shows our system produces consistent and complete summaries compared to human written ones.

cs.CL↗

arXiv2020

Traffic Agent Trajectory Prediction Using Social Convolution and Attention Mechanism

The trajectory prediction is significant for the decision-making of autonomous driving vehicles. In this paper, we propose a model to predict the trajectories of target agents around an autonomous vehicle. The main idea of our method is considering the history trajectories of the target agent and the influence of surrounding agents on the target agent. To this end, we encode the target agent history trajectories as an attention mask and construct a social map to encode the interactive relationship between the target agent and its surrounding agents. Given a trajectory sequence, the LSTM networks are firstly utilized to extract the features for all agents, based on which the attention mask and social map are formed. Then, the attention mask and social map are fused to get the fusion feature map, which is processed by the social convolution to obtain a fusion feature representation. Finally, this fusion feature is taken as the input of a variable-length LSTM to predict the trajectory of the target agent. We note that the variable-length LSTM enables our model to handle the case that the number of agents in the sensing scope is highly dynamic in traffic scenes. To verify the effectiveness of our method, we widely compare with several methods on a public dataset, achieving a 20% error decrease. In addition, the model satisfies the real-time requirement with the 32 fps.

cs.LG↗

arXiv2020

A Super-resolution Optical Classifier with High Photon Efficiency

We propose and demonstrate a photon-efficient optical classifier to overcome the Rayleigh limit in spatial resolution. It utilizes mode-selective sum frequency generation and single-pixel photon detection to resolve closely spaced incoherent sources based on photon counting statistics. Super-resolving and photon efficient, this technique can find applications in microscopy, light detection and ranging (LiDAR), and astrophysics.

quant-ph↗

arXiv2020

Wide-field, high-resolution lensless on-chip microscopy via near-field blind ptychographic modulation

We report a novel lensless on-chip microscopy platform based on near-field blind ptychographic modulation. In this platform, we place a thin diffuser in between the object and the image sensor for light wave modulation. By blindly scanning the unknown diffuser to different x-y positions, we acquire a sequence of modulated intensity images for quantitative object recovery. Different from previous ptychographic implementations, we employ a unit magnification configuration with a Fresnel number of ~50,000, which is orders of magnitude higher than previous ptychographic setups. The unit magnification configuration allows us to have the entire sensor area, 6.4 mm by 4.6 mm, as the imaging field of view. The ultra-high Fresnel number enables us to directly recover the positional shift of the diffuser in the phase retrieval process, addressing the positioning accuracy issue plagued in regular ptychographic experiments. In our implementation, we use a low-cost, DIY scanning stage to perform blind diffuser modulation. Precise mechanical scanning that is critical in conventional ptychography experiments is no longer needed in our setup. We further employ an up-sampling phase retrieval scheme to bypass the resolution limit set by the imager pixel size and demonstrate a half-pitch resolution of 0.78 micron. We validate the imaging performance via in vitro cell cultures, transparent and stained tissue sections, and a thick biological sample. We show that the recovered quantitative phase map can be used to perform effective cell segmentation of the dense yeast culture. We also demonstrate 3D digital refocusing of the thick biological sample based on the recovered wavefront. The reported platform provides a cost-effective and turnkey solution for large field-of-view, high-resolution, and quantitative on-chip microscopy.

eess.IV↗

arXiv2020

LinearPartition: Linear-Time Approximation of RNA Folding Partition Function and Base Pairing Probabilities

RNA secondary structure prediction is widely used to understand RNA function. Recently, there has been a shift away from the classical minimum free energy (MFE) methods to partition function-based methods that account for folding ensembles and can therefore estimate structure and base pair probabilities. However, the classical partition function algorithm scales cubically with sequence length, and is therefore a slow calculation for long sequences. This slowness is even more severe than cubic-time MFE-based methods due to a larger constant factor in runtime. Inspired by the success of our recently proposed LinearFold algorithm that predicts the approximate MFE structure in linear time, we design a similar linear-time heuristic algorithm, LinearPartition, to approximate the partition function and base pairing probabilities, which is shown to be orders of magnitude faster than Vienna RNAfold and CONTRAfold (e.g., 2.5 days vs. 1.3 minutes on a sequence with length 32,753 nt). More interestingly, the resulting base pairing probabilities are even better correlated with the ground truth structures. LinearPartition also leads to a small accuracy improvement when used for downstream structure prediction on families with the longest length sequences (16S and 23S rRNA), as well as a substantial improvement on long-distance base pairs (500+ nt apart).

q-bio.BM↗