Rumelhart, Hinton & Williams, Learning representations by back-propagating errors, Nature 1986 · Eq. 1–9
The Forward–Backward Chain Rule Lab
Back-propagation in a 2-2-1 network, one micro-step at a time
The network
inputs, hidden units, output · each unit shows x | y
i₁
i₂
d
Play
speed
Run full epoch
Train 100 epochs
epoch
0
· E
–
Parameters
value · gradient · change
Randomize weights
Reset
Learning
Eq. 8 and 9
learning rate ε
momentum α
when to change the weights
after each case
after each sweep
All cases
output now, and E over the epochs