Speaker |
Ciara Pike-Burke |
|---|---|
Affiliation |
Imperial College London |
Date |
Friday, 21 February 2025 |
Time |
12:00-13:00 |
Location |
UCL Centre for Artificial Intelligence, 1st Floor, 90 High Holborn, London WC1V 6BH |
Link |
https://ucl.zoom.us/j/99748820264 |
Event series |
Jump Trading/ELLIS CSML Seminar Series |
Abstract |
In this talk, we explore the foundations of goal-conditioned hierarchical reinforcement learning (HRL). Hierarchical reinforcement learning decomposes complex decision-making problems into manageable sub-tasks, offering the potential for more efficient learning. However, the effectiveness of this approach depends on the particular hierarchal decomposition considered. We derive a lower bound on the sample complexity of HRL, which provides a criterion for determining when a hierarchical decomposition can be beneficial. In settings where our lower bound shows that considering hierarchy could lead to improved performance, we derive efficient hierarchical algorithms that do indeed enjoy this performance gain. |
Biography |
Ciara is a lecturer in statistics at Imperial College London. Her research is in statistical machine learning. She is interested in sequential decision making problems under uncertainty and potentially limited feedback. In particular, she works on multi-armed bandit, reinforcement learning and online learning problems. |