LIRa session (online only): Andrés Occhipinti Liberman

Please, notice, that this talk has been RESCHEDULED, from 19th November to 26th November.

Date and Time: Thursday, November 26th 2020, 16:30-18:00, Amsterdam time.

Venue: online.

Title: Learning to Act and Observe in Partially Observable Domains


We consider a learning agent in a partially observable domain (or environment), with which the agent has never interacted before. The agent wishes to learn a representation of the domain dynamics: how the agent’s actions affect the state of the domain and what becomes observable as a result of such actions. To produce such knowledge, the learner has access to experience gathered by taking actions in the domain and observing their results.
We present algorithms for learning “as much as possible” (in a well-defined sense) both about what is directly observable and about what actions do in the domain, given the learner’s observational constraints. We differentiate the level of domain knowledge attained by each algorithm, and characterize the type of observations required to reach it. The algorithms use dynamic epistemic logic (DEL) to represent the learned domain information symbolically. Our work continues that of Bolander and Gierasimczuk (2015), which developed DEL-based learning algorithms based to learn domain information in fully observable domains.
This is joint work with Thomas Bolander and Nina Gierasimczuk.