Attention-Gated Reinforcement Learning of Internal Representations for
Classification
Pieter R. Roelfsema & Arjen van Ooyen, Neural Computation (2005) 17:
2176-2214.
Abstract
Animal learning is associated with changes in the efficacy of
connections between neurons. The rules that govern this plasticity can
be tested in neural networks. Rules that train neural networks to map
stimuli onto outputs are given by supervised learning and reinforcement
learning theories. Supervised learning is efficient but biologically
implausible. In contrast, reinforcement learning is biologically
plausible but comparatively inefficient. It lacks a mechanism that can
identify units at early processing levels that play a decisive role in
the stimulus-response mapping. Here we show that this so-called credit
assignment problem can be solved by a new role for attention in
learning. There are two factors in our new learning scheme that
determine synaptic plasticity: (1) a reinforcement signal that is
homogeneous across the network and depends on the amount of reward
obtained after a trial, and (2) an attentional feedback signal from the
output layer that limits plasticity to those units at earlier processing
levels that are crucial for the stimulus-response mapping. The new
scheme is called attention-gated reinforcement learning (AGREL). We show
that it is as efficient as supervised learning in classification tasks.
AGREL is biologically realistic and integrates the role of feedback
connections, attention effects, synaptic plasticity, and reinforcement
learning signals into a coherent framework.
For full text, go to
http://www.bio.vu.nl/enf/vanooyen/papers/agrel2005_abstract.html
-- Dr. Arjen van Ooyen Center for Neurogenomics and Cognitive Research (CNCR) Department of Experimental Neurophysiology Vrije Universiteit De Boelelaan 1085 1081 HV Amsterdam The NetherlandsE-mail: arjen.van.ooyen@falw.vu.nl Phone: +31.20.5987090 Fax: +31.20.5987112 Room: B329 Web: http://www.bio.vu.nl/enf/vanooyen
|
Home Login Meetings Courses Belgium Maillist Credits Servers Links Archive
Please send comments and suggestions to
Mike Wijnants. |
|
Page last updated on © BBF 1998 all rights reserved |