桑基韬
博士、教授、计算机科学系主任
博士、教授、计算机科学系主任
办公电话:010-51688055 | 电子邮件: jtsang@bjtu.edu.cn |
通讯地址:北京交通大学九教北506 | 邮编:100044 |
桑基韬,出生于山东烟台。东南大学学士、中科院自动化研究所博士。2012-2017年在模式识别国家重点实验室工作,2017年入选北京交通大学“卓越百人”计划。曾获中科院院长特别奖、中科院百篇优博、CCF优秀博士论文提名、ACM中国新星奖等。主要研究方向为多媒体内容分析、网络数据挖掘、可信赖机器学习等。已出版英文专著一部,第一/第二作者的相关工作7次获得中国计算机学会推荐国际会议的论文奖项。国家高层次青年人才计划入选者,作为负责人先后主持国家自然科学基金重点项目、国家重点研发计划(首批新一代人工智能重大项目)课题、北京市杰出青年基金等,以第二完成人获得中国电子学会自然科学一等奖和北京市科学技术奖。
2017.05 – 现在 北京交通大学 四级教授、三级教授
2012.07 – 2017.04 中国科学院自动化研究所 助理研究员、副研究员
2015.11 – 2016.04 微软亚洲研究院 铸星计划访问学者
2012.09 – 2014.07 中国-新加坡数字媒体研究院(中科院与新加坡国立大学联合实验室) Adjunct Researcher
研究主页:
https://adam-bjtu.org/ (2017-)
https://www.nlpr.ia.ac.cn/mmc/homepage/jtsang.html (-2017)
公众号:"ADaM应用数据挖掘和机器学习"
【教学】
招收计算机科学与技术专业方向的硕士(保研+统考,学硕+专硕)和博士生(直博+硕博一体化+普博)。欢迎具有扎实数学或编程基础、有志于机器学习理论研究或应用机器学习/数据挖掘方法解决实际问题的同学联系我。
-Revisiting Visual Model Robustness: A Frequency Long-Tailed Distribution View. NeurIPS 2023
-Adaptive Adversarial Logits Pairing. ACM ToMM, 2023.
-Unlearnable Clusters: Towards Label-Agnostic Unlearnable Examples. CVPR 2023.
-Attention, Please! Adversarial Defense via Activation Rectification and Preservation. ACM ToMM 2023
-Towards Adversarial Attack on Vision-Language Pre-training Models. ACM Multimedia 2022
-Investigating and Explaining the Frequency Bias in Image Classification. IJCAI 2022
-Robust CAPTCHAs Towards Malicious OCR. TMM 2021.
公平性和算法去偏
-Fair Visual Recognition via Intervention with Proxy Features. ACM Multimedia 2023
-Unsupervised Debiasing via Pseudo-bias Labeling in an Echo Chamber. ACM Multimedia 2023
-Debiasing backdoor attack: A benign application of backdoor attack in eliminating data bias. Information Sciences 2023.
-Towards Alleviating the Object Bias in
Prompt Tuning-based Factual Knowledge Extraction. ACL
Findings 2023.
-Counterexample Contrastive Learning for Spurious Correlation Elimination. ACM Multimedia 2022
-Counterfactually Measuring and Eliminating Social Bias in Vision-Language Pre-training Models. ACM Multimedia 2022
-Towards Accuracy-Fairness Paradox: Adversarial Example-based Data Augmentation for Visual Debiasing. ACM Multimedia 2020
可解释性
-TIF: Threshold Interception and Fusion for Compact and Fine-grained Visual Attribution. TMM, accepted.
-杨朋波,桑基韬等. 面向图像分类的深度模型可解释性研究综述。软件学报,2023.
-基于特征归因重要性评价的卷积网络剪枝。CCDM 2020 (最佳学生论文)
数据隐私保护
-Unlearnable Clusters: Towards Label-Agnostic Unlearnable Examples. CVPR 2023.
-JPEG Compression-Resistant Low-Mid Adversarial Perturbation against Unauthorized Face Recognition System. Information Sciences, accepted.
-Benign Adversarial Attack: Tricking Models for Goodness. ACM Multimedia 2022
-Adversarial
privacy-preserving filter, ACM
Multimedia 2020 .
多模态基础模型
-From Association to Generation: Text-only Captioning by Unsupervised Cross-modal Mapping. IJCAI 2023.
-Improved Visual Fine-tuning with Natural Language Supervision. ICCV 2023.
-Towards Adversarial Attack on Vision-Language Pre-training Models. ACM Multimedia 2022
-Counterfactually Measuring and Eliminating Social Bias in Vision-Language Pre-training Models. ACM Multimedia 2022
推荐系统和用户建模
-Knowledge Graph-enhanced Sampling for Conversational Recommender System. TKDE, accepted.
-Image-Based Personality Questionnaire Design. ToMM 2022.
-Learning to Learn a Cold-start Sequential Recommender. TOIS 2022.
授权专利: