Reward Optimization in the Primate Brain: A Probabilistic Model of Decision Making under Uncertainty
Figure 3
Relationship between Model and Neural Activity.
The input to the model is a random dots motion sequence. Neurons in MT with tuning curves emit
spikes at time step
, which constitutes the observation
in the POMDP model. The animal maintains the belief state
by computing
(
can be parameterized by
and
- see text). The optimal policy is implemented by selecting rightward eye movement
when
, or equivalently, when
(and likewise for leftward eye movement
).