I would like to nominate Eric Buehler in the category “Spot Talent from the Field”. His WSRP 2025 project “Training and exploring the inner workings of discrete neural networks” – for which I was a mentor – has the goal to “reverse engineer” Artificial Neural Networks (ANN) by using discrete mathematics. (Instead of the traditional ANN back-propagation continuous mathematics.)
The project started with the motivation and theoretical background of Stephen Wolfram’s work “What’s Really Going On in Machine Learning? Some Minimal Models”. Eric developed a package extending the original code, and designed, staged, and conducted different discrete-ANN experiments. (Based on hexagonal cellular automata.) For those experiments he utilized different optimization algorithms (simulated annealing being the winner) and programmed the orchestration of parallel executions of the experiments.
The reported results are very promising – ANN over certain data can be done using discrete mathematics. Extensions of Eric’s work would be of great interest!