epoch 0 | train
A B

what to notice

The gentlest curve in the catalogue: watch the straight opening boundary discover its first bend. If you only watch one thing, watch the loss curve's elbow - almost all of the improvement happens in the first tenth of training.

instruments

loss

log scale
– train -- test

gradient magnitude by layer

Gradients shrink as they travel back – the products-of-slopes fact from Maths of Learning ch3, live.

permalink

learning rate0.03