Hands-On Q-Learning with Python

Baseline models in Q-learning and machine learning research

In machine learning terminology, we refer to the randomly-acting agent that we have created in this chapter as a baseline model. A baseline is intended to give us a comparison metric to improve against. We use baseline models in machine learning to validate the performance of our current models against our past models.

Using a baseline model is good practice in scientific research, because we want to build on the progress that others have made, along with our own progress. As researchers, we start with the best available models of a scientific phenomenon and show how we can improve on their predictive capability. In this way, we create value for the scientific community by building ...