⬅️ Simulator: Backpropagation
∂L/∂w = (∂L/∂y) · (∂y/∂z) · (∂z/∂w)
ERROR PROPAGATION DIRECTION (REVERSE FLOW ⟵)
Conv1 Kernels
∂L/∂K
⟵
Max-Pooling
∂L/∂X_pool
⟵
FC Layers
∂L/∂W_fc
⟵
Loss / Softmax
∂L/∂y_pred
🔗 Chain Rule in Current Layer:
Description: Figure 9.8: Interactive Backpropagation simulator: advance through the Chain Rule steps to follow the error signal flow in the reverse direction of the network, observing the computation of partial derivatives of the gradient at each layer.