| Michael Herrmann |
| Thu 27 Feb 2020, 12:45 - 14:00 |
| IF, G.03 |
If you have a question about this talk, please contact: Jodie Cameron (jcamero9)
Title: Information Theory for Autonomous Agents
Abstract: The talk will discuss the role of information in probabilistic robotics. Information theory is based on probabilities that typically are unknown to agents in a new environment. The estimation of probabilities by sampling would require the agent to perform long-lasting repetitive behaviour over a learning phase in order to achieve stationarity and independence. We consider the triple learning task that the agent faces when it is to execute a task in an environment: Exploration, planning, and solution of a task (alternatively: sampling, control, and optimisation). This context suggests an information-theoretic approach to the exploration-exploitation problem, which we will discuss in the talk at three levels: An outline of the main ideas, highlights of the formalism, and a discussion of a few examples.