Abstract : We consider a biologically plausible model of the basal gan-glia that is able to learn a probabilistic two armed bandit task using reinforcement learning. This model is able to choose the best option and to reach optimal performances after only a few trials. However, we show in this study that the influence of exogenous factors such as stimuli salience and/or timing seems to prevail over optimal decision making, hence questioning the very definition of action-selection. What are the ecological conditions for optimal action selection ?
https://hal.inria.fr/hal-01333210
Contributor : Nicolas P. Rougier
<>
Submitted on : Friday, June 17, 2016 - 8:41:46 AM
Last modification on : Friday, May 10, 2019 - 5:31:14 PM
Long-term archiving on : Sunday, September 18, 2016 - 10:51:43 AM
Bhargav Teja Nallapu, Nicolas P. Rougier. Dynamics of reward based decision making a computational study. ICANN 2016 , Sep 2016, Barcelona, France. ⟨hal-01333210⟩
