"""Learning — the parametric policy (policy.py) and the closed loop (loop.py)
that improves it from simulated outcomes. See ../SIMULATION.md."""