| |
Abstract:
Recent evidence suggests that dopaminergic neurons in
vertebrates report prediction errors during reward based classical
and instrumental conditioning. We consider more complicated
conditioning paradigms which involve combining predictions from
multiple predictive stimuli. We show that our existing model fails
to act in accordance with the learning data, and suggest an
alternative in which there is attentional selection between
different available stimuli. The new model is a form of mixture of
experts (Jacobs, Jordan & Barto, 1991) and is statistically
well-founded.
|
