Skip to content
Reinforcement Learning

Function approximation

Learning objective

Section titled “Learning objective”

Drop the lookup table. Generalise across states so the method survives problems too large to enumerate.

Planned scope

Section titled “Planned scope”
  • Why tabular methods stop working
  • Linear function approximation and feature construction
  • Tile coding and coarse coding
  • Semi-gradient TD and the semi-gradient trick
  • The deadly triad: bootstrapping, off-policy, approximation

Prerequisites

Section titled “Prerequisites”

Everything above this chapter in the sidebar.