| Title: |
Neural mechanisms of instrumental learning : neuroimaging, pharmacological and stimulation studies in humans ; Les mécanismes neuraux de l'apprentissage par renforcement : études en neuroimagerie, pharmacologique et de la stimulation chez l'homme |
| Authors: |
Skvortsova, Vasilisa |
| Contributors: |
Paris 6; Pessiglione, Mathias; Plassmann, Hilke |
| Source: |
Theses.fr |
| Publication Year: |
2015 |
| Subject Terms: |
Récompense; Punition; Effort; Dopamine; L'apprentissage par renforcement; Noyau pédonculopontin; Reward; Punishment; psy; edu |
| Description: |
Learning from actions is a key ability for survival. But do we learn differently depending on whether the action outcome is positive or negative? Did the brain integrate different choice dimensions such as rewards, punishments or physical efforts in the same way? Do they all rely on the same neural circuit? Does dopamine influence both learning from rewards and efforts? Reinforcement learning theory postulates that learning follows stepwise minimization of the difference between prediction (e.g. internal representation of expected outcome) and actual outcome. We investigated how brain activity relates to these internal variables in different types of learning and how these representations are altered by pharmacological manipulation and deep brain stimulation.In study 1, we found an increase in power in beta band (10-20Hz) in response to reward in the peduncolopontine nucleus (PPN) of patients with Parkinson’s disease. Stimulation of the PPN specifically improved learning from rewards but not from punishments. This brainstem structure might contribute to the reward-related representations in the midbrain dopamine neurons that are known for their computations of reward prediction errors.In the studies 2 and 3, we compared learning to maximize reward with learning to minimize effort. FMRI results suggest that reward and effort related computations are carried by partially dissociable neural networks. Moreover, dopamine, a neuromodulator known to enhance reward maximization did not influence learning to minimize efforts.Overall, this PhD helps generalizing learning algorithms across different choice dimensions and specifying their implementation in different neural networks. ; Savoir apprendre de ses actions est crucial pour la survie de l'individu. Apprenons-nous différemment selon que nos actions sont récompensées ou punies? Le cerveau intègre-t-il de la même façon les différentes dimensions du choix, tels que les récompenses, les punitions et l'effort physique? Ces dimensions de choix sont-elles représentées par ... |
| Document Type: |
thesis |
| Language: |
English |
| Relation: |
10670/1.oc44h3; http://www.theses.fr/2015PA066297 |
| Availability: |
http://www.theses.fr/2015PA066297 |
| Rights: |
other |
| Accession Number: |
edsbas.7676B23D |
| Database: |
BASE |