Tsetlin Machines as an Alternative to Neural Networks
Neural networks dominate much of today’s AI landscape, from image generation to language models. But they are not the only way to build useful machine learning systems. One interesting alternative is the Tsetlin Machine (TM), introduced by Ole-Christoffer Granmo and colleagues at the University of Agder in Norway. Tsetlin Machines are especially interesting if you care about: interpretability (understanding why a model makes a decision) compact models for edge devices logic-style rule learning rather than purely numeric function approximation This post gives a practical overview of what a Tsetlin Machine is, how it works, and where it may fit as an alternative to neural networks. At the end, I discuss what we know so far about use in arts and creativity. ...