Email updates

Keep up to date with the latest news and content from BMC Neuroscience and BioMed Central.

This article is part of the supplement: Sixteenth Annual Computational Neuroscience Meeting: CNS*2007

Open Access Poster presentation

Learning sensitivity derivative by implicit supervision

Mohamed N Abdelghani1*, Timothy P Lillicrap4 and Douglas B Tweed123

Author Affiliations

1 Department of Physiology University of Toronto, Toronto, Ontario M5S 1A8, Canada

2 Department of Medicine, University of Toronto, Toronto, Ontario M5S 1A8, Canada

3 Centre for Vision Research, York University, Toronto, Ontario M3J 1P3, Canada

4 Centre for Neuroscience Studies, Queen's University, Kingston, Ontario K7L 3N6, Canada

For all author emails, please log on.

BMC Neuroscience 2007, 8(Suppl 2):P201  doi:10.1186/1471-2202-8-S2-P201

The electronic version of this article is the complete one and can be found online at:

Published:6 July 2007

© 2007 Abdelghani et al; licensee BioMed Central Ltd.

Poster presentation

In control theory, variables called sensitivity derivatives quantify how a system's performance depends on the commands from its controller. Knowledge of these derivatives is a prerequisite for adaptive control, including sensorimotor learning in the brain, but no one has explained how the derivatives themselves could be learned by real neural networks, and some say they aren't learned at all but are known innately. Here we show that this knowledge can't be solely innate, given the adaptive flexibility of neural systems. And we show how it could be learned using forms of information transport available in the brain. The mechanism, which we call implicit supervision, explains how sensorimotor systems cope with high-dimensional workspaces, tools, and other task complexities. It accelerates learning and explains a wide range of findings on the limits of adaptability, which are inexplicable by any theory that relies solely on innate knowledge of the sensitivity derivatives.