What they Learn

A1

World

state

actions

reward

0-Level

A2

A3

A4

state

action

A1

A2

A3

A4

A1

A3

A1

A3

A2

A1

A2

A3

2-Level

1-Level

Not Feasible to imple-ment. Approximate?

Models others as

State->Action Probability Pairs

Reinforcement Learning

Previous slide

Next slide

Back to first slide

View graphic version