Baum-Welch Algorithm for Bottom Up Option Learning by Imitation

Barbieri, Davide (2019) Baum-Welch Algorithm for Bottom Up Option Learning by Imitation. Bachelor's Thesis, Artificial Intelligence.

Preview	Text AI_BA_2019_DAVIDEBARBIERI.pdf Download (757kB) | Preview
	Text Toestemming.pdf Restricted to Registered users only Download (139kB)

Abstract

The Baum-Welch algorithm (Jelinek, Bahl, and Mercer, 1975) was developed to fit a hidden Markov model that maximizes the likelihood of a given sequence of observations (with a given number of hidden states). Fox, Krishnan, Stoica, and Goldberg (2017) proposed to look at option learning by imitation as solving a hidden Markov model, where the observations are primitives in a given solution trajectory, and the options are hidden probability distributions over primitives generating such trajectory. This paper adapts the Baum-Welch algorithm to fit the above idea by accounting for the differences between an option hierarchy and a hidden Markov model. The results seem to suggest that the architecture is capable of learning by imitation a series of static options. It is also able to exploit them to solve previously unseen tasks, similar to the ones solved by the imitation data.

Item Type:	Thesis (Bachelor's Thesis)
Supervisor name:	Wiering, M.A.
Degree programme:	Artificial Intelligence
Thesis type:	Bachelor's Thesis
Language:	English
Date Deposited:	28 Aug 2019
Last Modified:	11 Sep 2019 06:52
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/20815

Actions (login required)

View Item