Reinforcement Learning
The aim of the coursework is to better familiarise you with function approximation and sampling in the context of Reinforcement Learning (RL).
Abstract
Data Healthy volunteers were scanned with simultaneous bolus+infusion [11C]raclopride positron emission tomography (PET) and functional magnetic resonance imaging (fMRI) to measure dopamine release and neural activity during a 20 minute differential fear conditioning paradigm, pairing one cue (CS+) with an aversive electrical shock while another cue (CS-) was never paired with shock. Skin conductance was recorded during the fear conditioning. Thus, the data consists of PET measure of changes in dopamine concentration (5 minute frames), fMRI measure of neural activity (3 s TR), and skin conductance responses during fear conditioning. Skin conductance responses (SCRs) to CS+ subtracted from CS-is used as the fear learning measure. Data collection is ongoing. We have so far scanned ten volunteers (meanSD age: 25.15.9 years; 7 women) and plan to scan 8 more. Question Functional neuroimaging has revealed a critical role for amygdala in acquisition of fear memories, and animal studies have indicated that amygdala dopamine signaling is necessary for fear conditioning. However, little is known regarding the relationship between dopamine signaling, amygdala activity and fear learning in humans. Thus, we are interested in testing these relationships in our dataset. We are aware that this is a bit vague, but are happy to discuss more specific hypothesis during the meeting.