🔄 Backpropagation Step-by-Step

How Backpropagation Works

Forward Pass: Input flows through network, calculating outputs layer by layer

Loss Calculation: Compare prediction with actual output, compute error

Backward Pass: Calculate gradients using chain rule, propagate error backward

Weight Update: Adjust weights using gradient descent: w = w - α∇L

Applications: Training all neural networks • The foundation of deep learning