The Bernstein Center Freiburg

Informal Seminar
Hansol Choi

Korea Advanced Institute of Science and Technology
Department of Electrical Engineering

Reward Related Actions using Hierarchical Models

There are increasing behavioral and neuroscientific reports that the reward related actions utilize internal models of the world. Several model-based decision methods, such as hierarchical reinforcement learning and planning as inference framework, has been suggested to model the process. The methods explain how the possible future states influence ongoing decisions. There are several problems in modeling the model-based decision process of brains. They include 1) how to model the statistical structure of the world, 2) how to recognize ongoing state using the model and 3) how to process the information to produce behaviors. I recognize that the first and second problems are fundamentally identical to the traditional perception problems. Iíll introduce reward-related behavior algorithms based on hierarchical models, which are derived from the models for perception systems.

Friday, May 16, 2014

12:15 h
BCF Library
First Floor
Hansastr. 9a
The talk is open to the public. Guests are cordially invited!