A neural network learning to classify 2D points in real-time. Watch signals flow forward, errors propagate back, and decision boundaries form.
Blue vs orange regions show how the network classifies each point in 2D space
Cross-entropy loss decreasing as the network learns the pattern