AI Security and Privacy Team Seminar (Talk by Lu Sun, School of Information Science and Technology, ShanghaiTech University ).

This is an online seminar. Registration is required.

【 AI Security and Privacy Team】
【Date】2024/February/29(Thu) AM 09:00-10:00 (JST)
【Speaker】Lu Sun
School of Information Science and Technology, ShanghaiTech University

Title: Personalized and Efficient Multi-Task Learning via Model Compression

Abstract:
Multi-Task Learning (MTL) is an emerging machine learning topic, which seeks to improve the generalization performance of multiple learning tasks by sharing common information across them. It has been applied in many real-world applications, such as computer vision, natural language processing, bioinformatics analysis and ubiquitous computing. The key challenge in MTL is how to capture task correlations among different tasks in a personalized and efficient way.
In this talk, I will give an overview of our research works on personalized and efficient MTL. First, I will present our work on optimization of classifier chains for multi-label classification (MLC), a special case of MTL. We consider to optimize classifier chains via conditional likelihood maximization, in either or both of two aspects: label correlation modeling and multi-label feature selection. Second, I will introduce our MTL work on how to handle multi-view data in a fast and robust way. To deal with task-view outliers in real problems, we propose to promote joint group-sparsity on decomposed feature parameters and achieve view consistency by group trace lasso. Finally, I will present our recent works on personalized MLC and efficient MTL. These works aim to (1) allow each sample has its own MLC model to improve generalization and interpretability, and (2) efficiently implement MTL by deep tensor/matrix factorization or structured sparse learning. We will also present experimental results that demonstrate the efficacy of the proposed methods.
References:
1. Weijia Lin, Jiankun Wang, Lu Sun, Mineichi Kudo and Keigo Kimura, “Multi-Label Personalized Classification via Exclusive Sparse Tensor Factorization”, in Proceedings of the 23rd IEEE International Conference on Data Mining (ICDM 2023), 398-407, 2023. DOI: 10.1109/ICDM58522.2023.00049
2. Luhuan Fei, Lu Sun, Mineichi Kudo and Keigo Kimura, “Structured Sparse Multi-Task Learning with Generalized Group Lasso”, in Proceedings of the 26th European Conference on Artificial Intelligence (ECAI 2023), 692-699, 2023. DOI: 10.3233/FAIA230333
3. Zijian Yang, Zhiwei Li and Lu Sun, “Generalized Discriminative Deep Non-Negative Matrix Factorization Based on Latent Feature and Basis Learning”, in Proceedings of the 32nd International Joint Conference on Artificial Intelligence (IJCAI 2023), 4486-4494, 2023. DOI: 10.24963/ijcai.2023/499
4. Lu Sun, Canh Hao Nguyen, Hiroshi Mamitsuka, “Fast and Robust Multi-View Multi-Task Learning via Group Sparsity”, in Proceedings of the 28th International Joint Conference on Artificial Intelligence (IJCAI 2019), 3499-3505, 2019. DOI: 10.24963/ijcai.2019/485
5. Lu Sun and Mineichi Kudo, “Optimization of Classifier Chains via Conditional Likelihood Maximization”, Pattern Recognition, 74: 503-517, 2018. DOI:10.1016/j.patcog.2017.09.034