School of
Information Technology and Electrical Engineering

Speaker: Dr Miao Xu
Seminar Date: Thu, 10/10/2019 - 11:00
Venue: 47A-470; Sir James Foots building
Host: Prof Amin Abbosh

Seminar Type: Guest Research Seminar

Abstract: 

Machine learning has been applied ubiquitously in the real world and a successful machine learning model is usually trained with large-scale and high-quality data. Although we can collect data in scales, a key challenge is that the collected data are often imperfect and complex. This challenge requires both robust learning algorithms and human involvement. In this talk, I will introduce my studies on interactive robust learning with complex data and my future plan. To enhance robustness on learning with complex data, we developed a novel matrix completion technique exploiting side information. We also developed an active feature acquisition method to interact with humans for saving the feature acquisition cost. Both theoretical analysis and experimental studies validated the effectiveness of our proposed methods. In the future, facing the imperfect, dynamical, and unstable real world, I plan to develop novel machine learning techniques for the well-being of people and society. My plan also includes building a trustworthy intelligent system, which can interact with humans and learn from a massive volume of complex data for healthcare applications.

Biography: 

Miao Xu is a postdoctoral researcher at the RIKEN Center for Advanced Intelligence Project, supervised by Prof. Masashi Sugiyama. She pursued her Ph.D. from the Department of Computer Science at Nanjing University, supervised by Prof. Zhi-Hua Zhou and was a visiting scholar at Michigan State University, supervised by Prof. Rong Jin. Her research focuses on robust machine learning and its real-world applications. She has published 15 papers on conferences, journals, and workshops, including ICML, NeurIPS, AAAI, IJCAI, KDD, and TKDE. She has served as program committees to ICML, NeurIPS, AISTATS, AAAI, IJCAI, and ACML. She received the NJU Presidential Fellowship, IBM Ph.D. Fellowship, and CAAI Outstanding Doctoral Dissertation Award.