XDXiangying Dang

Xiangying Dang

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Xiangying Dang has a total of 31 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Violation-Aware Mutation Testing for Mathematical Contracts of Deep Learning Primitives

    by Xiaolin Huang, Xiangying Dang, Fan Wang, Dunwei Gong

    CONTEXT: Deep Neural Networks (DNNs) deployed in safety-critical domains rely on computational primitives such as Softmax,LayerNorm, and MatMul whose numerical contracts govern inference correctness. When violated at lower-level primitives, deviationsaccumulate and amplify layer by layer, leading to inaccurate inference or catastrophic failures. However, existing deep learningmutation testing techniques (e.g., DeepMutation, DeepCrime) rely primarily on structural mutations and numerical perturbations,failing to reach internal numerical semantics. Furthermore, perturbations are readily absorbed by built-in stabilization mechanisms,rendering such faults invisible.OBJECTIVES: To lower mutation testing granularity from model and data levels to the primitive level, formalize numericalcontracts as test oracles, and establish a framework for auditing primitive-level numerical semantic correctness.METHODS: We formalize numerical contracts—encompassing probability normalization, statistical preservation, numericalboundedness, and others—as test oracles, and establish a four-layer hierarchical taxonomy of contract violations (L1–L4): numericalstability, statistical moments, distributional axioms, and structural invariants. We propose behavioral fingerprints using violation-trigger frequency vectors to replace binary kill matrices, and design a two-stage Hierarchical Violation-Aware Reduction (HVAR) toreduce scale while preserving diversity. Evaluation covers eight representative primitives.RESULTS: The fingerprint achieves a 38.1% resolution gain (RG = 1.381) and zero cross-layer merging (CM = 0) over traditionalkill matrices. HVAR reduces test scale by approximately 75% while outperforming global and random baselines in coverage lowerbounds. Critically, the framework reveals systematic contract violations in mutants surviving traditional tolerance oracles: allsurviving Softmax and LayerNorm mutants trigger cross-layer violations, and 100% of surviving RBF Kernel mutants violate L1numerical stability.CONCLUSION: This work extends mutation testing to the numerical propagation control of deep learning systems, providing aframework for primitive-level testing toward numerical semantic correctness, and advancing machine learning software testing fromfunctional correctness to numerical contract preservation.

  • Enhanced MGA-based test data generation via mutation-relevant domain partitioning

    by Shen Shen, Xiangying Dang, Guosheng Hao, Dunwei Gong, Xiangjuan Yao

  • Open Access

    Efficient Multiple Path Coverage in Mutation Testing with Fuzzy Clustering-Integrated MF_CNNpro_PSO

    by Qian Qu, Xiangying Dang, Heng Xia, Lei Tao

    Fault concealment in complex software programs and the difficulty of generating test cases to detect such faults present significant challenges in software testing. To resolve these challenges, this paper suggests a novel method that integrates mutation testing, fuzzy clustering, convolutional neural networks (CNN), and particle swarm optimization (PSO) to efficiently generate test cases that cover multiple paths with numerous faults (mutant branches). Initially, mutation-based paths are classified using fuzzy clustering based on their coverage difficulty and similarity. A multi-feature CNN model (MF_CNNpro) is then constructed and trained on the paths of each cluster. Finally, the predicted particles from the MF_CNNpro model are used as the initial population for PSO, which evolves to generate the test cases. The proposed method is evaluated on six test programs, and the results demonstrate that it significantly improves clustering separation and reduces clustering compactness. By selecting only the cluster center paths to construct the MF_CNNpro model, training and prediction costs are effectively reduced. Moreover, the use of MF_CNNpro and PSO to select representative individuals as the initial population greatly enhances the evolutionary efficiency of PSO. The proposed method outperforms traditional approaches in clustering, prediction, and test data generation. Specifically, the SC clustering method improves cluster separation (SP) by 0.021, reduces compactness (CP) by 0.054, and decreases clustering rate (CR) by 4.97%, thereby enhancing clustering precision. The MF_CNNpro model improves the IA metric by 38.2% and reduces the U-Statistic and MSE by 83.0% and 97.9%, respectively, optimizing prediction performance. The MF_CNNpro+PPSOpro method increases the path coverage success rate from 47.9% to 97.4% (a 103.3% improvement), reduces the number of iterations by 84.1%, and decreases execution time by 95.6%, significantly improving generation efficiency.

  • Optimizing test data generation using SI_CNNpro-enhanced MGA for mutation testing

    by Lei Tao, Xiangying Dang, Juxin Hu, Dunwei Gong, Guosheng Hao, Xiangjuan Yao, Bingsen Huang


  • Test data generation for covering mutation-based path using MGA for MPI program

    by Xiangying Dang, Jinyong Wang, Dunwei Gong, Xiangjuan Yao, Changqing Wei, Biao Xu


  • The Nodes Influence Maximization in Open Source Software Community Based on Probability Propagation Model

    by Qing Zhao, Xiangjuan Yao, Xiangying Dang, Dunwei Gong

    It is of great significance to effectively and quickly identify the most influential users in the open source software community, which can be represented by a complex network. Traditional methods of measuring node influence only consider some topology characteristics of the network, so the results are one-sided. Because there are different interactive behaviors among users in the community, it is necessary to mine more interaction information, so as to give a more comprehensive measure of node influence. In view of this, this paper proposes a method to maximize the node influence of open source software community based on a probability propagation model. Firstly, according to the relationship of users and their interactive records on projects, this paper quantifies the feedback of users on the project from three aspects (i.e. approve, save and modify) and establishes a new probability propagation model between users. Secondly, this paper proposes an algorithm(SIUF) to evaluate users' influence in the open source software community based on the probability propagation model. The algorithm fully considers the interaction behavior of users in the community. In the first stage, the user's own activity is taken as the initial ranking. In the second stage, the influence of neighbor nodes is taken into account, and the SIUF value of nodes is accumulated while the rich club effect is weakened. Finally, the proposed theory and method are applied to GitHub, a typical open source software community, and the correctness and effectiveness of this method are verified from influence spread and speed.


  • Enhancement of Mutation Testing via Fuzzy Clustering and Multi-Population Genetic Algorithm

    by Xiangying Dang, Dunwei Gong, Xiangjuan Yao, Tian Tian, Huai Liu

    Mutation testing, a fundamental software testing technique, which is a typical way to evaluate the adequacy of a test suite. In mutation testing, a set of mutants are generated by seeding the different classes of faults into a program under test. Test data shall be generated in the way that as many mutants can be killed as possible. Thanks to numerous tools to implement mutation testing for different languages, a huge amount of mutants are normally generated even for small-sized programs. However, a large number of mutants not only leads to a high cost of mutation testing, but also make the corresponding test data generation a non-trivial task. In this paper, we make use of intelligent technologies to improve the effectiveness and efficiency of mutation testing from two perspectives. A machine learning technique, namely fuzzy clustering, is applied to categorize mutants into different clusters. Then, a multi-population genetic algorithm via individual sharing is employed to generate test data for killing the mutants in different clusters in parallel when the problem of test data generation as an optimization one. A comprehensive framework, termed as $\mathbf {FUZGENMUT}$ , is thus developed to implement the proposed techniques. The experiments based on nine programs of various sizes show that fuzzy clustering can help to reduce the cost of mutation testing effectively, and that the multi-population genetic algorithm improves the efficiency of test data generation while delivering the high mutant-killing capability. The results clearly indicate that the huge potential of using intelligent technologies to enhance the efficacy and thus the practicality of mutation testing.