A5024100506- Software Testing and Debugging Techniques
- Natural Language Processing Techniques
- Topic Modeling
- Advanced Malware Detection Techniques
- Sentiment Analysis and Opinion Mining
- Software Engineering Research
- Adversarial Robustness in Machine Learning
- Medical Imaging Techniques and Applications
- Software Reliability and Analysis Research
- Text Readability and Simplification
- Software System Performance and Reliability
- Advanced Neural Network Applications
- Anomaly Detection Techniques and Applications
- Machine Learning and Algorithms
- Advanced Text Analysis Techniques
Beihang University
2019-2024
Hefei University of Technology
2024
Central Queensland University
2024
Swinburne University of Technology
2024
Understanding and predicting types of bugs are practical importance for developers to improve the testing efficiency take appropriate steps address in software releases. However, due complex conditions under which faults manifest complexity classification rules, automatic Mandelbugs is a difficult task. In this article, we present deep semantic information-based Mandelbug method that combines model with learning classifier makes use both labeled unlabeled bug reports. By training report on...
Recently, metamorphic testing (MT) has been used to augment test datasets by inclusion of new (follow-up) cases constructed from existing (source) using relations (MRs). It reported that the augmented usually have higher fault detection capabilities. is natural ask which contributes improvement To investigate this issue, we conducted an empirical study on three DNN models feeding 70,000 handwritten digits images, in six sets MRs were designed. We found follow-up better capabilities than...
Existing works have studied the impacts of order words within natural text. They usually analyze it by destroying original to create a scrambled sequence, and then comparing models' performance between sequences. The experimental results demonstrate marginal drops. Considering this findings, different hypothesis about word is proposed, including ``the redundant with lexical semantics'', ``models do not rely on order''. In paper, we revisit aforementioned hypotheses adding reconstruction...
<title>Abstract</title> Deep neural networks (DNNs) are now widely used in many sectors of our society. This phenomenon also means that if these DNNs contain faults, they will have profound adverse impacts on daily lives. Thus, to be comprehensively tested for ''correctness'' before released use.Since such testing involves the use a DNN test set, comprehensiveness this set is utmost importance. Until now, researchers proposed their own neuron-coverage (NC) metrics measure set. However,...
We study sentiment analysis task where the outcomes are mainly contributed by a few key elements of inputs. Motivated two-streams hypothesis, we explore processing input items and their weights separately developing neural architecture, named TraceNet, to address this type task. It not only learns discriminative representations for target via its encoders, but also traces at same time locators. In both encoders locators organized in layer-wise manner, smoothness regularization is employed...
Word order, a concept in linguistics essential for conveying accurate meaning, is seemingly not that necessary on language models based the existing works. Contrary to this prevailing notion, our paper delves into impacts of word order by employing carefully selected tasks demand distinct abilities. Through utilization three controllable perturbation strategies, one novel qualification metric, four well-chosen tasks, and languages, we conduct experiments shed light topic. Empirical findings...