Mila talk - Baihan Lin Unified Models of Human Behavioral

VIDEO DESCRIPTION:

Play, download and edit the free video Mila talk - Baihan Lin Unified Models of Human Behavioral Agents: Bandit, Contextual Bandit and RL.

Mila RL Sofa 07/03/2020
Invited talk by Baihan Lin from Columbia University.

Talk Title: Unified Models of Human Behavioral Agents: Bandits, Contextual Bandits, and RL

Abstract: Artificial behavioral agents are often evaluated based on their consistent behaviors and performance to take sequential actions in an environment to maximize some notion of cumulative reward. However, human decision making in real life usually involves different strategies and behavioral trajectories that lead to the same empirical outcome. Motivated by clinical literature of a wide range of neurological and psychiatric disorders, we propose here a more general and flexible parametric framework for sequential decision making that involves a two-stream reward processing mechanism. We demonstrated that this framework is flexible and unified enough to incorporate a family of problems spanning multi-armed bandits (MAB), contextual bandits (CB), and reinforcement learning (RL), which decompose the sequential decision-making process in different levels. Inspired by the known reward processing abnormalities of many mental disorders, our clinically-inspired agents demonstrated interesting behavioral trajectories and comparable performance on simulated tasks with particular reward distributions, a real-world dataset capturing human decision-making in gambling tasks, and the PacMan game across different reward stationarities in a lifelong learning setting. Moreover, from the behavioral modeling perspective, our parametric framework can be viewed as the first step towards a unifying computational model capturing reward processing abnormalities across multiple mental conditions and user preferences in long-term recommendation systems. The talk consists of results from:

A Story of Two Streams: Reinforcement Learning Models from Human Behavior and Neuropsychiatry (AAMAS 2020)
Baihan Lin, Guillermo Cecchi, Djallel Bouneffouf, Jenna Reinen, Irina Rish
arxiv.org/abs/1906.11286

Unified Models of Human Behavioral Agents in Bandits, Contextual Bandits and RL (arXiv)
Baihan Lin, Guillermo Cecchi, Djallel Bouneffouf, Jenna Reinen, Irina Rish
arxiv.org/abs/2005.04544

Online Learning in Iterated Prisoner's Dilemma to Mimic Human Behavior (submitted to NeurIPS 2020)
Baihan Lin, Djallel Bouneffouf, Guillermo Cecchi
arxiv.org/abs/2006.06580

Speaker Bio: Baihan Lin is a Ph.D. Candidate (2017-2022) in the Center for Theoretical Neuroscience and Zuckerman Mind Brain Behavior Institute at Columbia University pursuing Computational Neuroscience. Baihan graduated from the University of Washington, Seattle (UW), in the NIH-funded Computational Neuroscience Training Program with a B.S. in Applied & Computational Mathematics (2017), a B.A. in Psychology (2017) with Honors, and an M.S. in Applied Mathematics (2020). His current theoretical research interest lies in the intersection between geometric topology, representation theory, dynamical systems, and complex networks, with extensive application interest in multiscale biological or cognitive systems, especially in understanding the neural systems and the theory of neural networks as well as developing brain-inspired algorithms in reinforcement learning and computer vision domains. Before attending Columbia, he researched on various interesting problems spanning vision neuroscience, mathematical biology, genome sciences, protein design, and human-computer interaction. Industry-wise, he maintains close collaborations with IBM Research on artificial intelligence and Microsoft Research on computational neuroscience.

Download, play and edit free videos and free audios from Mila talk - Baihan Lin Unified Models of Human Behavioral Agents: Bandit, Contextual Bandit and RL using RedcoolMedia.net web apps