CS294-112 Assignment Solutions

Solutions to the fall 2018 edition of the assignments for CS294-112: Deep Reinforcement Learning

Based on the video lectures from the Fall 2017 edition of the course.

Installation

Starting code

To be able to run the code for each homework, please start by following the instructions from each homework's PDF (linked with each assignment below) about how to install the homework. As for setting up a Python environment, the provided cs294.yml file is a conda requirements file that you can use to create a conda environment using the following command:

conda env create -f cs294.yml

This will create a conda environment named cs294.

Assignment 1: Imitation Learning:

• Behavioral Cloning (Done)
• DAgger

Assignment 2: Policy Gradients:

• Computation Graph (Done)
• Policy Gradient (Done)
• Neural Network Baselines (Done)

Assignment 3: Q-Learning and Actor-Critic:

• Q-Learning
• Actor-Critic

Assignment 4: Model-Based Reinforcement Learning:

• Problem 1
• Problem 2
• Problem 3a
• Problem 3b