Benutzer: Gast  Login
Titel:

Localized active learning of Gaussian process state space models

Dokumenttyp:
Konferenzbeitrag
Art des Konferenzbeitrags:
Textbeitrag / Aufsatz
Autor(en):
A. Capone; G.Noske; J. Umlauft; T. Beckers; A. Lederer; S. Hirche
Abstract:
While most dynamic system exploration techniques aim to achieve a globally accurate model, this is generally unsuited for systems with unbounded state spaces. Furthermore, many applications do not require a globally accurate model, e.g., local stabilization tasks. In this paper, we propose an active learning strategy for Gaussian process state space models that aims to obtain an accurate model on a bounded subset of the state-action space. Our approach aims to maximize the mutual information...     »
Stichworte:
data_driven_control; coman
Herausgeber:
Proceedings of Machine Learning Research
Kongress- / Buchtitel:
Learning for Dynamics & Control
Jahr:
2020
Jahr / Monat:
2020-06
Monat:
Jun
Reviewed:
ja
 BibTeX