🎯 Recommendation System Bipartite Graph (2D)
A real bipartite user-item graph: random embeddings, cosine similarity scoring, a similarity threshold and a diversity-injection control that reshapes which edges light up.
This 2D companion replaces the 3D original's decorative firing-node animation with the recommender logic its title actually promises: random 6-D embeddings for every user and item, a cosine-similarity score per pair, and a threshold slider that decides which edges count as real recommendations, plus a diversity slider that deliberately surfaces a few below-threshold pairs.
Random embeddings, cosine similarity scoring, a similarity threshold and diversity-injection control reshape which user-item edges are recommended, live.
2D · HTML5 Canvas 2D · 60 FPS target · runs fully client-side, no install