Intelligent Systems

Explore the Context: Optimal Data Collection for Context-Conditional Dynamics Models

2021

Conference Paper

ev

In this paper, we learn dynamics models for parametrized families of dynamical systems with varying properties. The dynamics models are formulated as stochastic processes conditioned on a latent context variable which is inferred from observed transitions of the respective system. The probabilistic formulation allows us to compute an action sequence which, for a limited number of environment interactions, optimally explores the given system within the parametrized family. This is achieved by steering the system through transitions being most informative for the context variable. We demonstrate the effectiveness of our method for exploration on a non-linear toy-problem and two well-known reinforcement learning environments.

Author(s): Jan Achterhold and Joerg Stueckler
Book Title: Proc. of the 24th International Conference on Artificial Intelligence and Statistics (AISTATS)
Year: 2021
Department(s): Embodied Vision
Research Project(s): Learning for Model-Based Control and Planning
Bibtex Type: Conference Paper (inproceedings)
Paper Type: Conference
Note: preprint CoRR abs/2102.11394
State: Published
URL: http://proceedings.mlr.press/v130/achterhold21a.html
Links: Preprint
Project page
Attachments: Poster

BibTex 

@inproceedings{achterhold2021_explorethecontext,
  title = {Explore the Context: Optimal Data Collection for Context-Conditional Dynamics Models},
  author = {Achterhold, Jan and Stueckler, Joerg},
  booktitle = {Proc. of the 24th International Conference on Artificial Intelligence and Statistics (AISTATS)},
  year = {2021},
  note = {preprint CoRR abs/2102.11394},
  doi = {},
  url = {http://proceedings.mlr.press/v130/achterhold21a.html}
}