Reinforcement Learning - A Gentle Introduction and Industrial Application

Explore reinforcement learning through a 50-minute conference talk delivered by Dr. Christian Hidber at MLCon. Gain intuitive insights into how reinforcement learning algorithms function, illustrated through the analogy of a child learning a new game. Discover the process of translating real-world problems into reinforcement learning tasks and learn about the challenges of implementing such solutions in production across 7000 clients in 42 countries. Examine an industrial application focused on siphonic roof drainage systems for large buildings, where reinforcement learning reduced the fail rate of an existing supervised learning solution by over 70%. Delve into topics such as policy algorithms, neural networks, dynamic programming, hydraulic simulations, and network architecture. Understand the practical implications of reinforcement learning in solving complex problems without the need for extensive labeled datasets.

Introduction
Game example
Policy
Algorithm
The problem
The Neural Network
Gaprate
Diameter
Success rate
Dynamic programming
Hydroelectric roofing
Hydraulic simulation errors
Actions
Gamestate
Reward Function
Network Architecture
Results
Implementation
Wrap up
Outro