Deep Learning on Azure with Python: Reinforcement Learning

This machine learning course focuses on reinforcement learning and how it uses artificial intelligence to find the best possible solution to complex problems involving multiple decisions.

Use reinforcement learning for complex problem solving

Reinforcement learning acknowledges the multifaceted, multilevel nature of the problems we use machine learning to solve. These challenges might be viewed as a sequence, with each resolved challenge creating or limiting possibilities to solve the next. Framing these challenges as relational learning problems allows us to explore every potential path through a sequence of decisions. This allows artificial intelligence to determine the most effective or efficient solution to complex problems.

Reinforcement learning can be applied to neural networks used in deep learning, helping us to build more refined algorithms.

Explore dynamic programming algorithms and more

This course will give you an introduction to reinforcement learning using Python, in Microsoft Azure. You’ll learn how to frame relational learning problems. You’ll get an introduction to common relational learning algorithms, including dynamic programming algorithms and temporal difference learning. And you’ll discover Project Malmo – a platform for AI experimentation built in Minecraft.

Frame reinforcement learning problems in Azure with Python

By the end of this course, you will have developed a clear understanding of reinforcement learning techniques, and the relevant formal notation. You’ll then be able to apply these in Microsoft Azure Cognitive Services, using Python programming.

This machine learning and artificial intelligence course is designed for those who would like to learn more about deep learning. Basic knowledge of python programming would be advantageous, as would solid maths and computer science skills.

• Course Introduction
• About this Course
• What is Reinforcement Learning?
• Applications of Reinforcement Learning
• Comparisons To Machine Learning
• Elements of Reinforcement Learning
• CloudSwyft Hands-On Lab: RL Environments and Random Agent
• Wrapping Up the Week
• Introduction to Reinforcement Learning
• Bandits Framework
• Regret Minimisation
• Bridge to Reinforcement Learning
• CloudSwyft Hands-On Lab: Bandits
• Wrapping Up the Week
• The Reinforcement Learning Problem
• Agent and Environment Interface
• Markov Decision Process
• CloudSwyft Hands-On Lab 3
• Basics of Dynamic Programming
• Wrapping up the week
• Applying Dynamic Programming & Policy Evaluation
• CloudSwyft Hands-On Lab 4
• Temporal Difference Learning - Policy Evaluation
• Temporal Difference Learning - Policy Optimisation
• CloudSwyft Hands-On Lab 5
• Wrapping Up the Week
• Function Approximation and Deep Q-Learning
• Function Approximation
• CloudSwyft Hands-On Lab 6
• RL with Deep Neural Networks
• Extensions to Deep Q-Learning
• CloudSwyft Hands-On Lab 7
• Introduction to Policy Optimisation
• Wrapping Up the Week
• Policy Gradient and Actor Critic
• Likelihood Ratio Methods
• CloudSwyft Hands-On Lab 8
• Variance Reduction
• CloudSwyft Hands-On Lab 9
• Actor-Critic
• CloudSwyft Hands-On Lab 10
• Course Completion