• Online, Self-Paced
Course Description

Some problems are too complicated to describe to a computer and to solve with traditional algorithms, which is why reinforcement learning is useful. In this course, you will learn the fundamentals of reinforcement learning.

Learning Objectives

Introducing Reinforcement Learning

  • start the course
  • describe reinforcement learning and list some of the techniques that agents can use to learn
  • describe additive rewards and discounted rewards
  • describe passive learning
  • describe how to use direct utility estimation for passive learning and how to define the Bellman Equation in the context of reinforced learning
  • describe temporal difference learning and contrast it with direct utility estimation
  • describe active learning and contrast it with passive learning
  • describe exploration and exploitation in the context of active reinforced learning and describe some of the exploration policies used in learning algorithms

Q-learning Algorithm

  • define Q-learning for reinforced learning
  • describe the different parts used in Q-learning and how these can be implemented
  • describe on-policy and off-policy learning and the difference between the two
  • describe why lookup tables aren't ideal for most reinforced learning tasks and how to build some function approximations that can make these problems possible
  • describe how deep neural networks can be used to approximate q-value for given states in Q-learning

Practice: Q-learning

  • describe Q-learning and how to set up the algorithm for a particular problem

Framework Connections

The materials within this course focus on the NICE Framework Task, Knowledge, and Skill statements identified within the indicated NICE Framework component(s):

Specialty Areas

  • Software Development
  • Technology R&D

Feedback

If you would like to provide feedback for this course, please e-mail the NICCS SO at NICCS@hq.dhs.gov.