Interactive Backpropagation

Understand how neural networks learn from their mistakes. Watch the errors flow backwards and adjust the connections to improve future predictions.

0.50-0.400.300.80-0.200.100.60-0.50Input1.0Input0.5Input-0.2Hidden?Hidden?Output?T=0.80

Step Explorer

1

Forward Pass: Hidden Layer

2

Forward Pass: Output Layer

3

Loss Calculation

4

Backprop: Output Gradients

5

Backprop: Hidden Gradients

6

Weight Updates

Controls

0.100
1x

Data Point

x1
x2
x3
True Target (y)

Loss Curve

Start training to see loss...