The Trace Model

In the trace model, information comes from a set of independent traces.

Example:
In a markov decision process world, you learn (s,a,r)^T where the initial state is chosen from P(s) and each action is chosen according to the algorithm and the state is chosen from P(s'|s,a).

The μ Trace Model

The trace model except you choose P(s).

Trace model = generative model when the world is deterministic.