Soft Actor-Critic (SAC)

This repository contains the final project completed in the context of the Reinforcement Learning course at École Polytechnique

The project is based on the paper:

Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor
Tuomas Haarnoja, Aurick Zhou, Pieter Abbeel, Sergey Levine (2018)

It includes a theoretical study of maximum-entropy reinforcement learning, an implementation of soft policy iteration in a discrete setting, and empirical experiments with Soft Actor-Critic (SAC) in continuous control environments.

The work was concluded by an oral presentation delivered in front of a jury as part of the course evaluation.

---

Repository Structure

• Code/ — Personal implementations and experiments
• Report/ — Slides
• Research Paper/ — Original research article (Haarnoja et al., 2018)

---

Authors

Lucien Le Gall
Alexandre Zenou
Remi Moshfeghi

École polytechnique