Machine Learning Models: From Birth to serving the Real-World by Sunita Sarawagi

Location: faculty hall

Department of Computer Science and Automation
CSA Golden Jubilee Frontier Lecture Series

Speaker : Prof. Sunita Sarawagi
Institute Chair Professor
Computer Science and Engineering
Indian Institute of Technology, Bombay


Date : Wednesday, January 08, 2020

Time : 4:00 PM

Venue : Faculty Hall, Indian Institute of Science


A Machine Learning (ML) Model is born by tting a function around examples sampled from an unknown distribution, and is designed to generalize to other examples drawn from the  same distribution. The machine learning edice largely rests on this foundation. However, modern ML models trained for challenging tasks like object detection, speech recognition, and  language understanding require huge amounts of labeled data and leave a large carbon footprint. In return, a model once trained needs to serve many diverse real-world settings that do  not perfectly match its birth setting. In this talk, I will discuss current research on propping the ML edice against such shifting foundation. We will span over current props like  calibration, out-of-sample detection, domain adaptation, domain generalization, and robustness and reect on some futuristic topics.

Biography of the speaker

Prof. Sunita Sarawagi is Institute Chair Professor in Computer Science and Engineering at IIT-Bombay. Prof. Sarawagi received her B.Tech. in Computer Science from the Indian Institute  of Technology, Kharagpur in May 1991. She received her M.S. and Ph.D. in Computer Science from the University of California at Berkeley where she studied under Michael  Stonebraker.

Following her Ph.D., Sarawagi did stints at IBM Almaden Research Center as research scholar, Carnegie Mellon as visiting faculty, and joined IIT-Bombay in 1999. Between July 2014 and July 2016 Prof. Sarawagi was Visiting Scientist at Google Inc. in Mountain View where she worked on deep learning models for personalizing and diversifying YouTube and Play  recommendations, improving Duo’s conversation assistance engine, and extracting attributes of classes from the Knowledge Graph.

Among her many awards are IBM Faculty award (2003 and 2008); Fellow of the  Indian National Academy of Engineering (INAE) (2013); PAKDD Most Influential Paper Award 2014 for the paper: “Discriminative Methods for Multi-labeled Classification Shantanu Godbole and Sunita Sarawagi in PAKDD 2004”. Prof. Sarawagi has several patents to her name. They  include a patent for “Database System and Method Employing Data Cube Operator for Group-By Operations” and a patent for “Efficient evaluation of queries with mining predicates”.

The Infosys Prize 2019 in Engineering and Computer Science is awarded to Prof.Sunita Sarawagi for her research in databases, data mining, machine learning and natural language processing, and for important applications of these research techniques. The prize recognizes her pioneering work in developing information extraction techniques for unstructured data.

Host Faculty : Prof. Shalabh Bhatnagar


Scroll Up