1、基于机器学习的软件工程与图像视频分析的新问题、新方法研究报告人：荆晓远 jingxy_Pattern Recognition And Software EngineeringMachine Learning and its Aplication Laboratoryhttp:/ 近期工作介绍基于机器学习的图像视频分析技术研究Outline基于机器学习的软件工程技术研究1-1 近期工作介绍 : 软件工程方向软件工程方向聚焦于软件缺陷预测、软件工作量估计技术。软件缺陷预测：1. Dictionary Learning based Software Defect Prediction, Intern
2、ational Conference on Software Engineering (ICSE), 2014【 他引 40次，成功地将字典学习技术用于低维的缺陷数据分类 】2. Heterogeneous Cross-company Defect Prediction by Unified Metric Representation and CCA-based Transfer Learning, ACM Conference on Foundation of Software Engineering (FSE), 2015【 他引 33次，与香港科技大学计算机系研究组同时发现异质缺陷预测问
3、题 】3. An Improved SDA based Defect Prediction Framework for both Within-project and Cross-project Class-imbalance Problems, IEEE Transactions on Software Engineering (TSE), 2017. 【 该论文应邀在ICSE 2017上做 Journal First Paper(期刊精选论文 )报告 】4. On the Multiple Sources and Privacy Preservation Issues for Hetero
4、geneous Defect Prediction， IEEE Transactions on Software Engineering (TSE), 2018, 在线出版 . 【 被评价为这一方向上目前最好的结果 】5. Cross-Project and Within-Project Semisupervised Software Defect Prediction: A Unified Approach, IEEE Transactions on Reliability (TR), 2018.以及 CCF的 B类期刊 ASE、 IST、 STVR、 IET Software和 B类会议
5、ICSME等。1-1 近期工作介绍 : 软件工程方向软件工作量估计：1. Missing Data Imputation Based on Low-rank Recovery and Semi-supervised Regression for Software Effort Estimation, International Conference on Software Engineering (ICSE), 2016. 【 首次提出缺失工作量数据补偿问题，被评价为当时最好的结果、里程碑的工作 】2. Privacy Preserving via Interval Covering Base
6、d Subclass Division and Manifold Learning Based Bi-directional Obfuscation for Effort Estimation, IEEE/ACM Conference on Automated SoftwareEngineering (ASE), 2016. 【 首次提出工作量估计的隐私保护问题 】3. Software Effort Estimation Based on Open Source Projects: Case Study of Github, Information andSoftware Technolog
7、y (IST), 2017.4. Online Software Effort Estimation, ICSE短文录用 , 2018.软件工程方向聚焦于软件缺陷预测、软件工作量估计技术。1-2 近期工作介绍 : 图像视频分析方向图像视频分析方向聚焦于行人重识别、高维数据特征提取技术。行人重识别：1. Super-resolution Person Re-identification with Semi-coupled Low-rank Discriminant Dictionary Learning, CVPR, 2015 / IEEE Transactions on Image Proce
8、ssing (TIP), 2017. 【 他引 52次，首次提出超分辨率行人重识别问题 】2. Video-Based Person Re-Identification by Simultaneously Learning Intra-Video and Inter-Video Distance Metrics, IJCAI, 2016 / IEEE Transactions on Image Processing (TIP), 2018.3. Learning Heterogeneous Dictionary Pair with Feature Projection Matrix for P
9、edestrian Video Retrieval via Single Query Image, AAAI, 2017. 【 首次提出图像视频行人重识别问题 】4. Image to Video Person Re-identification by Learning Heterogeneous Dictionary Pair with Feature Projection Matrix, IEEE Transactions on Information Forensics and Security (TIFS), 2018. 5. Semi-supervised Cross-view Pr
10、ojection-based Dictionary Learning for Video-based Person Re-identification, IEEE Transactions on Circuits and Systems for Video Technology (TCSVT), 2017. 【 首次提出 半监督视频行人重识别问题 】1-2 近期工作介绍 : 图像视频分析方向图像视频分析方向聚焦于行人重识别、高维数据特征提取技术。高维数据特征提取和学习 （特别是多源异构大数据分析，发展了多视图学习技术）：1. Semi-supervised Multi-view Correla
11、tion Feature Learning with Application to Webpage Classification, AAAI, 2017. 【 在多视图、多光谱、多模态特征学习方面发表了系列研究论文 】2. Multi-kernel Low-rank Dictionary Pair Learning for Multiple Features based Image Classification, AAAI, 2017. 3. Discriminant Tensor Dictionary Learning with Neighbor Uncorrelation for Imag
12、e Set Based Classification, IJCAI, 2017.4. Multiset Feature Learning for Highly Imbalanced Data Classification, AAAI, 2017. 5. Multi-Label Dictionary Learning for Image Annotation. IEEE Transactions on Image Processing, 2016.6. Multi-view Low-rank Dictionary Learning for Image Classification, Patter
13、n Recognition, 2016.7. Multi-spectral Low-rank Structured Dictionary Learning for Face Recognition, Pattern Recognition, 2016.8. Uncorrelated Multi-set Feature Learning for Color Face Recognition, Pattern Recognition, 2016.9. Intra-View and Inter-View Supervised Correlation Analysis for Multi-View F
14、eature Learning, AAAI, 2014.10.Uncorrelated Multi-View Discrimination Dictionary Learning for Recognition, AAAI, 2014.2 近期工作介绍基于机器学习的图像视频分析技术研究Outline基于机器学习的软件工程技术研究不同的软件模块 之间存在一定 相似性 ，即一个软件模块可以由少量的其他模块近似地表示近年来，很多流行的机器学习分类方法在软件缺陷预测领域都得到了应用。但是，这些分类方法通常会面临一些困难，例如 类不平衡 （ class-imbalance）问题和误分类代价 （ misc
15、lassification cost）问题。稀疏表示、字典学习是图像视频高维数据分析热点技术， 能否适用于低 维 度量表示的 软件缺陷数据？能否通 过 构建 软 件缺陷数据的稀疏表示 解决 预测过 程中遇到的 问题 ?软件缺陷预测基于字典学习的软件缺陷预测 2-1 Xiao-Yuan Jing, Shi Ying, Zhi-Wu Zhang, Shanshan Wu, Jin Liu: Dictionary learning based software defect prediction. ICSE, 2014.我们第一次尝试在软件缺陷预测领域应用字典学习技术，通过软件度量特征挖掘，学习多字典和特征表示系数。为了增强模型分类能力：软件缺陷预测2-1 Xiao-Yuan Jing, Shi Ying, Zhi-Wu Zhang, Shanshan Wu, Jin Liu: Dictionary learning based software defect prediction. ICSE, 2014.误分类代价 问题类不平衡问题代价敏感技术PCA降维技术通过增加对第二类错误分类的惩罚来提高分类能力利用 PCA技术对每个类的子字典进行初始化，同时解决类不平衡问题
