New math art: Optimization — gradient descent, Adam, genetic algorithms. Six panels: Rastrigin loss landscape with SGD/Adam/RMSProp trajectories, Rosenbrock saddle-point traversal, learning rate schedules (cosine/warmup/cyclic), convexity theory + subgradients, training/validation loss curves, genetic algorithm population evolution. Blog (with the math): https://ai.jskitty.cat/blog.html #math #machinelearning #optimization #generativeart #nostr