My first foray into the Mario Kart series was Mario Kart 64 but I appreciate its predecessor, Super Mario Kart. It had its own learning curve but utilised the Mode 7 graphics brilliantly. But what if you could program a bot to play to based on neural networks and Q-learning? Enter MarIQ.
But first a quick primer.
A neural network is a series of algorithms that tries to recognise relationships in a set of data through a process that mimics how your brain works (hence the name “neural” as it pertains to neurons or nodes).
Q-learning is an algorithm that learns the quality of actions and uses what it has learnt to make decisions under certain circumstances. It’s classed as a model-free learning algorithm because it doesn’t need a model (hence the connotation “model-free”) of the environment it learns from.
MarIQ’s entire goal is to accurately predict how much reward it will receive by taking different actions and the better it gets at predicting that, the better it gets at choosing the action that will take it through the course.
Seth Bling, its creator, said it took 80 hours to learn one course.
You can read a manual on how to set up MarIQ to use yourself.