Han-Jia Ye @ LAMDA, NJU

Email: yehj [at] lamda.nju.edu.cn
               yehj [at] nju.edu.cn
               yhjyehanjia [at] gmail.com

• Han-Jia Ye, Hong-You Chen, De-Chuan Zhan, Wei-Lun Chao. Identifying and Compensating for Feature Deviation in Imbalanced Deep Learning. arXiv:2001.01385, 2020. [Paper]

We identify the over-fitting and feature divation phenomena for minor classes in deep long-tail learning, and propose a simple yet effective class-dependent temperature to compensate for such influence.

• Han-Jia Ye*, Hexiang Hu*, De-Chuan Zhan, Fei Sha. Learning Classifier Synthesis for Generalized Few-Shot Learning. arXiv:1906.02944, 2019. [Paper]

We investigate the problem of generalized few-shot learning and propose a learning framework that learns how to synthesize calibrated few-shot classifiers in addition to the multi-class classifiers of "head" classes.

• Han-Jia Ye, Su Lu, De-Chuan Zhan. Distilling Cross-Task Knowledge via Relationship Matching. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR'20), Seattle, Washington, 2020. To appear. [Paper]

To reuse the cross-task knowledge, we distill the comparison ability and the local classification ability of the embedding and the top-layer classifier from a teacher model, respectively.

• Han-Jia Ye, Hexiang Hu, De-Chuan Zhan, Fei Sha. Learning Embedding Adaptation for Few-Shot Learning. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR'20), Seattle, Washington, 2020. To appear. [Paper][code]

For few-shot learning, we employ a type of self-attention mechanism to transform the embeddings from task-agnostic to task-specific in both seen and unseen classes.

• Wei-Lun Chao*, Han-Jia Ye*, De-Chuan Zhan, Mark Campbell, Kilian Q. Weinberger. A Meta Understanding of Meta-Learning. In: The Adaptive and Multitask Learning (AMTL) 2019 Workshop, Long Beach, CA, 2019. [Paper] [ArXiv]

By rethinking meta-learning as a kind of supervised learning, we can borrow supervised learning tricks for the meta-learning paradigm.

• Han-Jia Ye, De-Chuan Zhan, Yuan Jiang, Zhi-Hua Zhou. Rectify Heterogeneous Models with Semantic Mapping. In: Proceedings of the 35th International Conference on Machine Learning (ICML'18), Stockholm, Sweden, 2018. Page: 5630-5639. [Paper][code]

The reusability and evovability of a model are anlayzed in this paper. The proposed framework generates meta features and reuses model across heterogeneous feature domains.

• Han-Jia Ye, Xiang-Rong Sheng, De-Chuan Zhan, Peng He. Distance Metric Facilitated Transportation between Heterogeneous Domains. In: Proceedings of the 27th International Joint Conference on Artificial Intelligence (IJCAI'18), Stockholm, Sweden, 2018. Page: 3012-3018. [Paper][code]

We deal with a specific problem for Optimal Transport Domain Adaptation. Our method extends the ability of OTDA to heterogeneous domains.

• Han-Jia Ye, De-Chuan Zhan, Xue-Min Si, Yuan Jiang. Learning Mahalanobis Distance Metric: Considering Instance Disturbance Helps. In: Proceedings of the 26th International Joint Conference on Artificial Intelligence (IJCAI'17), Melbourne, Australia, 2017. Page: 3315-3321. [Paper][code]

A robust Mahalanobis distance metric learning approach dealing with both instance and side-information uncertainty effectively.

• Han-Jia Ye, De-Chuan Zhan, Xiaolin Li, Zhen-Chuan Huang, Yuan Jiang. College Student Scholarships and Subsidies Granting: A Multi-Modal Multi-Label Approach. In: Proceedings of the IEEE International Conference on Data Mining (IEEE ICDM'16), Barcelona, Spain, 2016, Page: 559–568. [Paper][code]

A multi-modal and multi-label method dealing with real-world college student scholarships and subsidies granting task.

• Han-Jia Ye, De-Chuan Zhan, Xue-Min Si, Yuan Jiang, Zhi-Hua Zhou. What Makes Objects Similar: A Unified Multi-Metric Learning Approach. In: Advances in Neural Information Processing Systems 29 (NIPS'16), Barcelona, Spain, 2016, Page: 1235-1243. [Paper][code]

A unified multi-metric learning approach discovering various types of semantics under objects linkages. Besides, we provide a unified solver with theoretical guarantee.

• Han-Jia Ye, Xue-Min Si, De-Chuan Zhan, Yuan Jiang. Learning Feature Aware Metric. In: Proceedings of the 8th Asian Conference on Machine Learning (ACML'16), Hamilton, New Zealand, 2016, Page: 286–301. [Paper][code]

A fast decomposition strategy learning sparse/robust Mahalanobis distance metric.

• Han-Jia Ye, De-Chuan Zhan, Yuan Jiang. Instance Specific Metric Subspace Learning: A Bayesian Approach. In: Proceedings of the 30th AAAI Conference on Artificial Intelligence (AAAI'16), Phoenix, AZ, 2016, Page: 2272-2278. [Paper]

A Bayesian perspective of distance metric learning, which can infer metric inductively.

• Han-Jia Ye, De-Chuan Zhan, Yuan Miao, Yuan Jiang, Zhi-Hua Zhou. Rank Consistency based Multi-View Learning: A Privacy-Preserving Approach. In: Proceedings of the 24th ACM International Conference on Information and Knowledge Management (CIKM'15), Melbourne, Australia, 2015, Page: 991-1000. [Paper][code]

A novel rank consistency criterion is proposed for multi-view learning in a privacy-preserving scenario.

• Yang Yang, Han-Jia Ye, De-Chuan Zhan, Yuan Jiang. Auxiliary Information Regularized Machine for Multiple Modality Feature Learning. In: Proceedings of the 24th International Joint Conference on Artificial Intelligence (IJCAI'15), Buenos Aires, Argentina, 2015, Page: 1033-1039. [Paper]

Improve the prediction ability of cheap weak modal feature with the help of its strong counterpart.

• Han-Jia Ye, Xiang-Rong Sheng, De-Chuan Zhan. Few-shot learning with adaptively initialized task optimizer: a practical meta-learning approach. Machine Learning. 2020, Volume 109, pp 643–664. [Paper]

A practical meta-learning approach which efficiently adapts the task-specific initialization to an effective classifier.

• Xiu-Shen Wei*, Han-Jia Ye*, Xin Mu, Jianxin Wu, Chunhua Shen, Zhi-Hua Zhou. Multi-instance learning with emerging novel class. IEEE Transactions on Knowledge and Data Engineering. To appear. [Paper]

A local metric learning approach to deal with the emerging novel class in multi-instance learning tasks.

• Han-Jia Ye, De-Chuan Zhan, Nan Li, Yuan Jiang. Learning Multiple Local Metrics: Global Consideration Helps. IEEE Transactions on Pattern Analysis and Machine Intelligence. To appear. [Paper]

By learning local metrics based on the global one, we try to adaptively allocate local metrics for heterogeneous data.

• Han-Jia Ye, De-Chuan Zhan. Few-Shot Learning via Model Composition (in Chinese). In: SCIENTIA SINICA Informatics (中国科学：信息科学). April 2020, Volume 50, Issue 5. [Paper]

We propose to compose classifiers inspired by the closed form of the least square loss, which fits learning with limited training examples.

• Han-Jia Ye, De-Chuan Zhan, Yuan Jiang. Fast Generalization Rates for Distance Metric Learning. Machine Learning. February 2019, Volume 108, Issue 2, pp 267–295. [Paper]

Theoretical analysis of distance metric learning with fast generalization rate

• Han-Jia Ye, De-Chuan Zhan, Yuan Jiang, Zhi-Hua Zhou. What Makes Objects Similar: A Unified Multi-Metric Learning Approach. IEEE Transactions on Pattern Analysis and Machine Intelligence. May 2019, Volume 41, Issue 5, pp 1257-1270. [Paper][Supplementary]

This manuscript extends our NIPS work. The concept of semantic metric, generalization analysis, and deep extension are introduced to get a more general framework.