🎬 2D Neural Collaborative Filtering versus Matrix Factorisation: When a Learned Interaction Beats the Dot Product
User-by-item heat maps of the true preference probability, matrix factorisation and neural collaborative filtering, with held-out cells outlined, live loss curves against the Bayes limit and five independent numerical checks (finite-difference gradients, ln 2 baseline, Bayes loss by Monte Carlo, monotone descent, parameter counts).
⚙ Under the hood
User-by-item heat maps of the true preference probability, matrix factorisation and neural collaborative filtering, with held-out cells outlined, live loss curves against the Bayes limit and five independent numerical checks (finite-difference gradients, ln 2 baseline, Bayes loss by Monte Carlo, monotone descent, parameter counts).
Neural Collaborative FilteringMatrix FactorisationRecommender SystemsEmbeddingsImplicit FeedbackBackpropagationBayes ErrorMachine Learning
2D · HTML5 Canvas 2D · 60 FPS target · runs fully client-side, no install