Hands-On Reinforcement Learning for Games

Imagination and reasoning in RL

Something that we can observe from our own experience of learning is how imagination can benefit the learning process. Pure imagination is the stuff of deep abstract thoughts and dreams, often closer to a hallucination than any way to solve a real problem. Except, this same imagination can be used to span gaps in our understanding of knowledge and allow us to reason out possible solutions. Say that we are trying to solve the problem of putting a puzzle together, and all we have are three remaining, mostly black pieces, as shown in the following image:

Imagining what the three missing puzzle pieces may look like ...