Reinforcement
Learning

Agenda

• Review of Machine Learning
• Introduction
• History
• Demo - Playground
• Future

Review of Machine Learning

Review of Machine Learning

Introduction

When to use Reinforcement Learning

• Data for learning currently does not exist
• Or you don’t want to wait to accumulate it
  (because delay might be costly)
• Or the data may change rapidly causing the
  outcome to change more rapidly than a typical
  model refresh cycle can accommodate.

Introduction

Typical RL Problems

• Robotic Control
• AI Game Play

RL - How it Works

• Environment - all possible values and steps
• State - current values
• Reward - benefit from action
• Agent - RL Algorithm
• Policy - Solution or Steps to Maximize Reward

The RL Bible

Reinforcement Learning:
An Introduction

Richard S. Sutton and Andrew G. Barto

RL - History

• 1963 - Tic Tac Toe
• 1992 - Backgammon
• 1997 - Deep Blue - Chess
• 2013 - Deep Mind - Atari
• 2016 - AlphaGo - Go
• 2017 - OpenAI Bot - DOTA

Playground - OpenAI

• Install it from Github:
  https://github.com/openai/gym
• Install either PyTorch or Keras
• PLAY

CartPole with ACER

The Future
From MOBA to RTS

Starcraft ][

• Deepmind and Blizzard announce open support of Starcraft 2 as an AI research environment.
• PySC2, a Starcraft II machine learning environment by DeepMind in Python.
• OpenAI interface