Reinforcement Learning
Peter Dayan
In CR Gallistel, editor, Steven's Handbook of Experimental
Psychology New York, NY: Wiley.
Abstract
Reinforcement learning studies the prediction and control of events
of affective importance in terms of psychological and
neural rules for adaptation. Reinforcement learning began as a
marriage between ideas in mathematical behavioral psychology and
artificial intelligence, and now has links all the way from
neuromodulatory systems in vertebrates to engineering and
statistical theories of adaptive optimizing control. In this
chapter, we describe the basic theory underlying reinforcement
learning, and its links with neuroscience, psychology, statistics
and engineering.
pdf