Digit Recognition - Backpropagation Process

Input
-
Ready
-
Output Prediction
Parameters Update
Layer $\mathbf{w}$ (Weight) $\mathbf{b}$ (Bias) $\nabla$ (Avg Gradient)
L1---
L2---
L3---
Output---
Global Loss $C_i^{(n+1)}$
-
Avg Activation
-
Net Avg Gradient
-
Prediction
-