Evolutionary Learning vs. The Zerg: Training & Evolving Star Craft II AIs in Python - Alan Smith

Genetic algorithms, Q learning and deep Q learning can enable powerful AI models to evolve and “learn” how to maximize their performance in a simulated system. Blizzard’s Star Craft II client API enables developers to connect to the Star Craft world and launch their code into battle with human and other AI players. This fast paced and sophisticated real time strategy game environment provides an ideal simulated world to explore using artificial intelligence. In this session Alan will explain the core concepts of genetic algorithms, Q learning and deep Q learning. He will then demonstrate how genetic algorithms can be applied to evolve an optimal resource gathering strategy for a Star Craft battle. A strong resource pool can then be leveraged to deploy a fighting force to defend bases, and then seek out and destroy the enemy. Q learning and deep Q learning can be leveraged to develop an algorithm that can “lean” how to optimize strategies to conquer different enemies in a number of scenarios. Throughout the session Alan will share tips for developers wanting to explore artificial intelligence and machine learning development in Python or C#. May this battle bring us glory!