Stat Colloquium: Dr. Shayok Chakraborty

Title: Learning with Weak Supervision: Algorithms and Applications

Abstract: The widespread deployment of inexpensive sensors has resulted in the generation of enormous amounts of digital data in today’s world. This has expanded the possibilities of solving real-world problems using computational learning frameworks; however, annotating the data (with class labels) to induce a machine learning model has remained a fundamental challenge. Thus, developing intelligent machine learning algorithms under the constraint of weak manual supervision is a problem of immense practical importance. Active learning and domain adaptation (or transfer learning) are two methods to address this problem. Active learning algorithms automatically select the salient and exemplar samples from large amounts of unlabeled data; this tremendously reduces human annotation effort as only a few samples selected by the algorithm need to be annotated manually. Domain adaptation algorithms leverage abundant labeled data in a source domain to develop a model for a related target domain, where labeled data is scarce, under the constraint of a probability distribution difference between the two domains. These two learning paradigms are even more important for training deep neural networks, which have demonstrated commendable empirical performance, but require a large amount of labeled data. In this talk, I will present an overview of several research problems in weakly supervised machine learning that I am currently exploring in my research.