🧪 Optimizer Race: SGD vs Momentum vs Adam on a Small Neural Network
Train the same small neural network three ways at once, from identical starting weights and identical mini-batches: plain SGD, momentum and Adam. Watch each decision boundary form on 2D datasets, compare loss curves and test accuracy, and verify backpropagation against finite differences.
⚙ Under the hood
Train the same small neural network three ways at once, from identical starting weights and identical mini-batches: plain SGD, momentum and Adam. Watch each decision boundary form on 2D datasets, compare loss curves and test accuracy, and verify backpropagation against finite differences.
Neural NetworksOptimizersBackpropagationAdamDecision BoundaryAI Algorithms
2D · HTML5 Canvas 2D · 60 FPS target · runs fully client-side, no install