The latest installment of @fluketc studies on how dopamine (DA) shapes learning rdcu.be/c3Ibf This was a tricky one; it combines a few ideas that together argue for a new way to understand DA function and account for data that evaded prior models. 1/n
6
30
109
26K
34
The main ideas, which we hope are helpful pointers to the full paper: Idea 1: DA modulates learning rate. Idea 2: Naive learning is policy search. Idea 3: DA encodes policy inputs and outputs Idea 4: DA manipulations should be referenced to endogenous signals 2/n
@jtdudman @dennis_a_burke @fluketc Congrats Luke and Josh! Have been following this one for a while, very cool story!
