The course explores alternatives to Reinforcement Learning for real-world problems by introducing Contextual Bandits and Imitation Learning. The learning outcomes include understanding the formal problems, limitations, and applications of these approaches. The course teaches skills such as using tools like Microsoft Azure and AWS SageMaker, and the methods of collecting trajectories, training models, and utilizing interactive experts. The intended audience is individuals interested in agent-based learning and seeking practical solutions for real-world challenges.
Alternatives to Reinforcement Learning for Real-World Problems
Overview
Syllabus
Intro
LET'S TALK ABOUT REINFORCEMENT LEARNING
THE THREE MACHINE LEARNS
EMBODIED LEARNING
AGENT-BASED LEARNING
THE DECISION POLICY
THE REWARD
TWO IDEAS
DEALING WITH UNCERTAINTY
REQUIREMENTS OF BIG SUCCESSES
SIMULATION
FULLY OBSERVABLE
TRANSFERABILITY OF METHOD
WHAT IS THE COST OF AN ERROR?
CAN WE APPLY THIS TO REAL PROBLEMS?
REAL-WORLD ALTERNATIVES
WHAT ARE WE TRYING TO SOLVE
TOOLS
MICROSOFT AZURE
AWS SAGEMAKER
WHEN SHOULD I USE CONTEXTUAL BANDITS?
LIMITATIONS
BEHAVIORAL CLONING
EXPERT SYSTEMS SUPERVISED LEARNING
COLLECT TRAJECTORIES FROM AN EXPERT
BREAK UP INTO STATE / ACTION PAIRS
TRAIN A MODEL ON THE TRAJECTORIES
INTERACTIVE EXPERTS
APPLICATIONS
WHEN SHOULD I USE IMITATION LEARNING?
SCALABILITY CONCERNS
CAPTURING DATASETS
IMITATION LEARNING + REINFORCEMENT LEARNING
RESOURCES
OFFLINE RL
WHY IS THIS EXCITING?
Taught by
Open Data Science
