University of Waterloo

Agent Autonomy via Self-supervision and Reinforcement Learning

Amin Oji, PhD Candidate

Dec 2, 2022, 11:30am EC4-2101A

This presentation will provide an overview of the plans for my PhD research to develop a self-supervised agent autonomy framework. Inspired by recent advances in self-supervised visual representation learning, vision Transformers, and model-based agent autonomy frameworks such as the World Models and Imagination-Augmented Agents (I2A), I will go over the background and the rationale for my proposed work, which aims to more closely mirror the ecology of perception in intelligent biological beings.