Can XCS learn without Genetic Algorithm using only covering?

What is it about?

XCSF is a rule-based machine learning technique. It learns a problem by training a set of rules, called a population. It does this by combining interactive learning, in the form of reinforcement learning, and evolution-inspired learning, in the form of a genetic algorithm. This paper investigates How do XCSF's genetic algorithm and rule generation contribute to the performance of the maze and deterministic frozen lake problems when using linear prediction or neural network-based prediction? Can the system learn the problems without the genetic algorithm being activated?

Featured Image

Why is it important?

XCS(F) still struggles to learn mutli-step Reinforcement Learning problems. Further insights in how its Genetic Algorithm effects this, might help to improve this in the future.