Skip to content

AEmanuelli/RL_internship

Repository files navigation

Multi-Agent Reinforcement Learning with Social Influence and Confirmation Bias in Music Listening

Welcome to the GitHub repository for my internship project at LNC2 of the DEC of ENS ULM, supervised by Mr. Stefano Vrizzi. In this repository, you will find the code and data related to my work.

Project Overview

This internship project, conducted one day per week, focused on multi-agent reinforcement learning with an emphasis on social influence and confirmation bias in the context of music listening. The project was divided into three main phases:

  1. Familiarization: To establish a foundational understanding of reinforcement learning concepts, I implemented the Rescorla-Wagner model for a single agent.

  2. Modeling Multiple Agents: Building upon the fundamentals, I extended the project to model multiple agents influencing each other's decisions, delving into the dynamics of social influence.

  3. Reproduction of Empirical Data: Leveraging the developed model, I sought to replicate empirical results from this research article. This exploration aimed to understand the impact of social influence on musical preferences.

Repository Contents

  • Code: The "classes.py" file in this directory contains Python classes I implemented for the project. These classes are designed to model the agents and their interactions. The alexis_.py file contains the code used for the last part, data repriduction

  • Data_Social_Influence: This directory contains data extracted from the article, which was used in an attempt to fit our model to the empirical data.

  • Figures_and_their_data: For transparency and reference, this directory holds the heatmaps presented in the report, along with the arrays of values computed to generate them.

  • RL_internship.ipynb: The Jupyter notebook included in this repository contains all the necessary code to obtain results from the models I implemented.

  • Presentation Slides: The final presentation slides are available in the "Présentation final stage.pdf" file. These slides provide a comprehensive overview of the project and its findings, with the presentation delivered in French.

About

Repository for my reinforcement Learning internship supervised by Stefano Vrizzi

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published