University of Cambridge > > CUED Control Group Seminars > Non-Stationary Representation Learning in Sequential Linear Bandits

Non-Stationary Representation Learning in Sequential Linear Bandits

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact Xiaodong Cheng.

Humans are naturally endowed with the ability to learn and transfer experience to later unseen tasks. One of the key mechanisms enabling such versatility is the abstraction of past experience into a ‘basis set’ of simpler representations that can be used to construct new strategies much more efficiently in future complex environments. What can we learn from humans when we design decision-making strategies? In this talk, we will talk about representation learning for multi-task decision-making in nonstationary environments. We consider the framework of sequential linear bandits, where the agent performs a series of tasks drawn from different environments. The embeddings of tasks in each environment share a low-dimensional feature extractor called representation, and representations are different across environments. We propose an online algorithm that facilitates efficient decision-making by learning and transferring non-stationary representations in an adaptive fashion.

This talk is part of the CUED Control Group Seminars series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.


© 2006-2024, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity