Thomas Akam

Thomas Akam

@thomasakam.bsky.social

Neuroscientist interested in planning and flexible behaviour | Open-source tool developer

148 followers132 following7 posts

Thomas Akam
@thomasakam.bsky.social

Why does dopamine ramp up during approach to predictable rewards? In this preprint with Luke Priestley, we explore the idea that dopamine ramps occur when reward predictions inferred using a world-model are used to train striatal cached values.

biorxiv.orgDopamine ramps as a normative consequence of dual-process controlMidbrain dopamine neurons are thought to implement a temporal difference (TD) reward prediction error (RPE) that updates cached values stored in striatum. This has been challenged by evidence that dop...
View full thread
Thomas Akam
@thomasakam.bsky.social

Why does dopamine ramp up during approach to predictable rewards? In this preprint with Luke Priestley, we explore the idea that dopamine ramps occur when reward predictions inferred using a world-model are used to train striatal cached values.

biorxiv.orgDopamine ramps as a normative consequence of dual-process controlMidbrain dopamine neurons are thought to implement a temporal difference (TD) reward prediction error (RPE) that updates cached values stored in striatum. This has been challenged by evidence that dop...
Thomas Akam
@thomasakam.bsky.social

The ramps are reward prediction errors, but the error is between what the world-model knows about future reward, and values already stored in striatum. The model explains diverse experimental data, and has implications for the nature of model-based evaluation and its interaction with TD learning.

You are up to date.