Skip Navigation


CIT can broadcast your seminar, conference or meeting live to a world-wide audience over the Internet as a real-time streaming video. The event can be recorded and made available for viewers to watch at their convenience as an on-demand video or a downloadable file. CIT can also broadcast NIH-only or HHS-only content.

Learning to distinguish the good from the bad: Neural systems underlying reinforcement learning

Loading video...

296 Views  
   
Air date: Friday, October 27, 2017, 12:00:00 PM
Time displayed is Eastern Time, Washington DC Local
Views: Total views: 296, (96 Live, 200 On-demand)
Category: NIH Director's Seminars
Runtime: 01:01:05
Description: Director's Seminar Series

The section on Learning and Decision making studies the neural circuitry and computational mechanisms that underlie reinforcement learning. Reinforcement learning (RL) is the behavioral process of learning what we like and do not like from experience. While some preferences are innate, many are learned over time. There are few areas in systems neuroscience where a theory has linked a neural mechanism to behavior as successfully as RL. The standard model of RL, developed in the mid-1990s, assumes that dopamine neurons code reward prediction errors (RPEs). RPEs are the difference between the reward that is received and the reward that was predicted. These dopaminergic RPEs are then communicated to the basal ganglia, specifically the striatum, because of its substantial dopamine innervation. The striatal neurons integrate RPEs to develop representations of the values of choices, i.e. our preferences. In contrast to the standard model we have recently shown that the amygdala, a structure in the medial temporal lobe, has a larger role in RL than the ventral striatum (VS). In addition, the role of the VS depends on the reward environment. When rewards are predictable, the VS has almost no role in learning whereas when rewards are less predictable the VS plays a larger role. These data outline a more specific role for the VS in RL than is attributed to it by current models. Given that the VS has been implicated in depression, particularly adolescent depression, this delineation of the contribution of the VS to normal behavior may help inform hypotheses about the mechanisms and circuitry underlying depression.
Debug: Show Debug
NLM Title: Learning to distinguish the good from the bad : neural systems underlying reinforcement learning / Bruno Averbeck.
Author: Averbeck, Bruno.
National Institutes of Health (U.S.),
Publisher:
Abstract: (CIT): Director's Seminar Series The section on Learning and Decision making studies the neural circuitry and computational mechanisms that underlie reinforcement learning. Reinforcement learning (RL) is the behavioral process of learning what we like and do not like from experience. While some preferences are innate, many are learned over time. There are few areas in systems neuroscience where a theory has linked a neural mechanism to behavior as successfully as RL. The standard model of RL, developed in the mid-1990s, assumes that dopamine neurons code reward prediction errors (RPEs). RPEs are the difference between the reward that is received and the reward that was predicted. These dopaminergic RPEs are then communicated to the basal ganglia, specifically the striatum, because of its substantial dopamine innervation. The striatal neurons integrate RPEs to develop representations of the values of choices, i.e. our preferences. In contrast to the standard model we have recently shown that the amygdala, a structure in the medial temporal lobe, has a larger role in RL than the ventral striatum (VS). In addition, the role of the VS depends on the reward environment. When rewards are predictable, the VS has almost no role in learning whereas when rewards are less predictable the VS plays a larger role. These data outline a more specific role for the VS in RL than is attributed to it by current models. Given that the VS has been implicated in depression, particularly adolescent depression, this delineation of the contribution of the VS to normal behavior may help inform hypotheses about the mechanisms and circuitry underlying depression.
Subjects: Amygdala--physiology
Choice Behavior--physiology
Dopaminergic Neurons--physiology
Learning--physiology
Reinforcement (Psychology)
Ventral Striatum--physiology
Publication Types: Lecture
Webcasts
Download: To download this event, select one of the available bitrates:
[64k]  [150k]  [240k]  [440k]  [740k]  [1040k]  [1240k]  [1440k]  [1840k]    How to download a Videocast
Caption Text: Download Caption File
NLM Classification: WL 337
NLM ID: 101716409
CIT Live ID: 26527
Permanent link: https://videocast.nih.gov/launch.asp?23554