Distributed AI with Ray Labs

The following labs are intended to give a basic introduction to the Ray distributed programming framework and examples of its use for 3 AI algorithms. This includes distributed implementations of Value Iteration for Markov Decision Process Planning, table-based Reinforcement Learning, and DQN for Deep Reinforcement Learning.

Direct any feedback to Alan Fern.

Lab 1: Introduction to the Ray Framework and Intel DevCloud

This lab introduces the basic Ray framework for distributed programming.

· Lab Instructions

· ray_tutorial.ipynb

· ray_tutorial.py

· map_reduce.py

Lab 2: Distributed Value Iteration for MDP Planning

This lab develops a distributed implementation of the Value Iteration algorithm for MDP planning.

· lab2.ipynb

Lab 3: Distributed Table-Based Reinforcement Learning

This lab develops and compares distributed implementations of the table-based reinforcement learning algorithms: Q-learning and SARSA.

· lab3.ipynb

Lab 4: Distributed Deep Reinforcement Learning

This lab develops a distributed implementation of the DQN algorithm for deep reinforcement learning.

· lab4.zip