Neural Network Training Visualization (4-Layer)

Training: Thickness = Update Magnitude. Stopped: Thickness = Final Weight Magnitude. Hidden Layers use ReLU.