Skip to content
Reinforcement Learning

Deep Q-networks

Learning objective

Section titled “Learning objective”

Neural networks as value approximators, and the machinery required to make that stable.

Planned scope

Section titled “Planned scope”
  • From linear features to a neural QQ
  • Experience replay
  • Target networks
  • Double DQN and duelling architectures
  • What actually makes DQN unstable

Prerequisites

Section titled “Prerequisites”

Everything above this chapter in the sidebar.