Speaker: Sheng Li, Data Scientist, Adobe Research, San Jose, CA

When: 8:30AM ~ 9:20AM, March 8, 2018

Where: CSB130

Abstract:High-dimensional data are ubiquitous in real-world applications, arising in images, videos, documents, online transactions, biomedical measurements, etc. Although data analytics in high-dimensional space is generally intractable due to the “curse of dimensionality”, significant progress has been made by exploiting the low-dimensional manifolds in high-dimensional data. Extracting effective and compact feature representations from high-dimensional data becomes a critical problem in data science and machine learning. Traditional data analytics methods, especially the statistical models, often make strong assumptions on the data distributions. However, real-world data might be contaminated by noise or captured from multiple views. Such uncertainty would hinder the performance of data analytics. In this talk, I will introduce some examples of my work in advancing the robust representation learning for data analytics, including: 1) low-rank and sparse modeling for robust graph construction; 2) multi-view learning for time-series classification; 3) subspace learning for causal inference. I will conclude this talk by describing my future research plan in the interdisciplinary field of data science.

Bio:Sheng Li is a Research Scientist at Adobe Research. He received his Ph.D. in Computer Engineering from the Northeastern University (NEU) in 2017. He has a broad interest in data science and machine learning, including multi-view learning, causal inference, deep learning, time series modeling, transfer learning and visual intelligence. He has published over 50 papers at leading conferences and journals, including NIPS, KDD, IJCAI, AAAI, ICCV, SIGIR, IEEE TPAMI/TIP/TKDE/TNNLS/TCSVT, ACM TKDD/TOMM, etc. He received three best paper awards (or nominations) at SDM 2014, ICME 2014 and IEEE FG 2013, and received the 2015 NEU’s Outstanding Graduate Student Research Award. He serves on the Editorial Board of four journals, and serves as a program committee member for several conferences such as IJCAI, AAAI and KDD. More could be found at: https://sites.google.com/view/shengli.