This repository contains various computational models designed to solve reinforcement learning problems.
This folder contains two models, one which fits data from a credit assignment task according to a Rescorla-Wagner learning rule, and the other according to an Eligibility Trace learning rule (see Sutton & Barto, 1998). Also included is an automated learning script which allows for parameter recovery testing as well as simulation of data according to either model. These models use maximum likelihood estimation to generate reinforcement learning parameters, specifically using fmincon.
This folder contains three models to fit data from a two-arm bandit task which delivers both positive and negative outcomes:
- Rescorla-Wagner ("symmetric") model
- Asymmetric Prediction Error model
- Asymmetric Outcome model