Productionizing Deep Reinforcement Learning with Spark and MLflow

Explore the practical application of Deep Reinforcement Learning (RL) in industry through this 26-minute talk from Databricks. Learn how Zynga leverages RL to personalize mobile games for millions of users daily. Discover the challenges and solutions in productionizing Deep RL applications using tools like Spark, MLflow, and TensorFlow. Gain insights on applying cutting-edge AI techniques to real-world scenarios, including tips for training RL agents and overcoming production challenges. Understand how to formulate personalization problems, design actions, choose appropriate RL algorithms, and implement automated hyperparameter tuning. Walk away with key takeaways on harnessing the power of Deep RL for business applications and improving user engagement at scale.

Syllabus

Intro
Game Design is Hard
Personalization Problem Formulation
Personalization Method 2: Prediction Models
Personalization Wishlist
Solution: Reinforcement Learning (RL)
RL Model Training
Academic RL Applications
Production RL Applications for Personalization
Production RL Challenges
RL-Bakery
Real Time Model Serving
Choose the Right Application
Designing Actions
Choosing RL Algorithms
Hyperparameter Tuning Automation
Key Takeaways