Machine learning visualisations

Visualisations About

All visualisations/Deep learning

How does error travel backward through a neural network?

Difficulty: Technical
Estimated time: 7 min

Loading visualisation

Step 1 of 4Set the evidence

Change the two inputs and inspect the weighted connections into the hidden layer.

Each hidden unit receives the same inputs but combines them with different authored starting weights.

Forward activations and backward gradients through a small neural network

Prediction 0.570, target 1, loss 0.561. 0 updates applied.

Input x10.80Input x20.20Learning rate0.50