experiments.md

Experiments

Find below the code that should be run on the command line using main.py for the different experiments.

A. Estimator Combination and Multi-level Abstraction

run a simple q-learning agent on the Willemsen gridworld as a baseline for improvements

python main.py A SIMPLE_Q WILLEMSEN

run a multi-estimator q-learning agent on the Willemsen gridworld to observe the differences in behavior

python main.py A ME_Q WILLEMSEN

B. Novelty as an Intrinsic Reward

to be defined

C. Guiding Exploration with Multi-level Abstractions on Novelty

to be defined