Reward Optimization in the Primate Brain: A Probabilistic Model of Decision Making under Uncertainty
Figure 1
POMP Framework for Decision Making.
Left: The graphical model representing the probabilistic relationship between random variables ,
,
and
. In the POMDP model, the hidden state
corresponds to coherence
and direction
jointly. The observation
corresponds to MT response
. The relations between these variables are summarized in table 1. Right: In order to solve a POMDP problem, the animal maintains a belief
, which is a posterior probability distribution over hidden states
of the world given observations
. At a current belief state
, an action is selected according to the learned policy
, which maps belief states to actions.