ConfPad

AlphaGo’s victory over the Go’s world champion was viewed dubiously by some critics as hype by a one-trick pony. Yet AlphaZero’s ability to learn chess in 4 hours and beat the strongest computer using not-of-this-world moves has silenced the strongest of sceptics. Reinforcement Learning is the cornerstone of building game-playing agents and along with algorithms such as Monte Carlo Tree Search (MCTS), they provide the tooling for building your own game-playing agent. This is exactly what we will go through in this talk: from representing the game rules to designing the network, we will cover how to build an agent to play and win Danske Bank's Hexagon (https://playhexagon.com/), a round-based strategy game. The talk is divided into a one-third introduction, history and basic theory, and two-thirds nitty-gritty of actually building the agent. Knowledge of Machine Learning basics is advantageous but the talk will be both informative and entertaining for the novice.