Abstract
Hutter’s optimal universal but incomputable AIXI agent models the environment as an initially unknown probability distribution-computing program. Once the latter is found through (incomputable) exhaustive search, classical planning yields an optimal policy. Here we reverse the roles of agent and environment by assuming a computable optimal policy realizable as a program mapping histories to actions. This assumption is powerful for two reasons: (1) The environment need not be probabilistically computable, which allows for dealing with truly stochastic environments, (2) All candidate policies are computable. In stochastic settings, our novel method Optimal Direct Policy Search (ODPS) identifies the best policy by direct universal search in the space of all possible computable policies. Unlike AIXI, it is computable, model-free, and does not require planning. We show that ODPS is optimal in the sense that its reward converges to the reward of the optimal policy in a very broad class of partially observable stochastic environments.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
Similar content being viewed by others
References
Hutter, M.: Universal Artificial Intelligence: Sequential Decisions based on Algorithmic Probability. Springer, Berlin (2004)
Levin, L.: Universal sequential search problems. Problems of Information Transmission 9(3), 265–266 (1973)
Schaul, T., Schmidhuber, J.: Towards Practical Universal Search. In: Proceedings of the Third Conference on Artificial General Intelligence, Lugano (2010)
Schmidhuber, J.: Sequential decision making based on direct search (Lecture Notes on AI 1828). In: Sun, R., Giles, C.L. (eds.) IJCAI-WS 1999. LNCS (LNAI), vol. 1828, p. 213. Springer, Heidelberg (2001)
Schmidhuber, J.: Optimal Ordered Problem Solver. Machine Learning 54, 211–254 (2004)
Schmidhuber, J.: Gödel machines: Fully Self-Referential Optimal Universal Self-Improvers. In: Goertzel, B., Pennachin, C. (eds.) Artificial General Intelligence, pp. 119–226 (2006)
Schmidhuber, J.: Ultimate Cognition à la Gödel. Cognitive Computation 1(2), 177–193 (2009)
Schultz, W., Dayan, P., Montague, P.R.: A neural substrate of prediction and reward. Science 275(5306), 1593 (1997)
Veness, J., Ng, K.S., Hutter, M., Silver, D.: A Monte Carlo AIXI Approximation. Technical Report 0909.0801, arXiv (2009)
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2011 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Glasmachers, T., Schmidhuber, J. (2011). Optimal Direct Policy Search. In: Schmidhuber, J., Thórisson, K.R., Looks, M. (eds) Artificial General Intelligence. AGI 2011. Lecture Notes in Computer Science(), vol 6830. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-22887-2_6
Download citation
DOI: https://doi.org/10.1007/978-3-642-22887-2_6
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-642-22886-5
Online ISBN: 978-3-642-22887-2
eBook Packages: Computer ScienceComputer Science (R0)