- This event has passed.
Seminar Series: Kiante Brantley
Please join us in welcoming Kiante Brantley.
Abstract
Natural language processing (NLP) systems that learn from feedback are essential to help real-world users with tasks. However, most NLP models are built from static data, unable to learn from feedback, but are deployed in environments where feedback signals exist. In this talk, I will focus on ideas for incorporating interactive machine learning into NLP. First, I will describe a new interactive machine-learning framework that combines large language models in NLP with modern reinforcement learning techniques. In the second part of the talk, I will describe a new framework for studying language-conditioned reinforcement learning in visual environments.
Biography
Kianté Brantley is a Postdoctoral scholar at Cornell working with Thorsten Joachims. He completed his Ph.D. in computer science at the University of Maryland College Park (UMD) advised by Professor Hal Daumé III. Brantley designs algorithms that efficiently integrate domain knowledge into sequential decision-making problems. He is most excited about imitation learning and interactive learning—or, more broadly, settings that involve a feedback loop between a machine learning agent decisions and the input the machine learning agent sees.