A two-step algorithm for learning from unspecific reinforcement

DOI10.1088/0305-4470/32/31/301MaRDI QIDQ4947717zbMATH OpenOpenAlexFDO

Authors R. Kühn, Ion-Olimpiu Stamatescu

Publication date 25 April 2000

Published in Journal of Physics A: Mathematical and General (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/cond-mat/9902354

zbMATH Keywords

Hebb rulelearning modelneural network

Mathematics Subject Classification ID

Neural nets applied to problems in time-dependent statistical mechanics (82C32) Neural networks for/in biological studies, artificial life and related topics (92B20)

Abstract: We study a simple learning model based on the Hebb rule to cope with "delayed", unspecific reinforcement. In spite of the unspecific nature of the information-feedback, convergence to asymptotically perfect generalization is observed, with a rate depending, however, in a non- universal way on learning parameters. Asymptotic convergence can be as fast as that of Hebbian learning, but may be slower. Moreover, for a certain range of parameter settings, it depends on initial conditions whether the system can reach the regime of asymptotically perfect generalization, or rather approaches a stationary state of poor generalization.

Recommendations

• Learning structured data from unspecific reinforcement
• Learning with incomplete information and the mathematical structure behind it
• Simple statistical gradient-following algorithms for connectionist reinforcement learning
• Hebbian errors in learning: an analysis using the Oja model
• Combining Hebbian and reinforcement learning in a minibrain model

Cited in

(2)

• Learning with incomplete information and the mathematical structure behind it
• Learning with incomplete information in the committee machine

This page was built for publication: A two-step algorithm for learning from unspecific reinforcement

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4947717)