The Backpropagation algorithm
The backprop algorithm for supervised learning
from exemplars is:
Repeat:

Send in inputs I_{i} for this exemplar.

Calculate outputs y_{k}

Get given correct outputs O_{k}

Measure E.
 w_{jk} weights:

Calculate all the
=
y_{k} ( 1  y_{k} ) ( y_{k}  O_{k} )

Calculate all the
=
y_{j}

For all j,k:
 w_{ij} weights:

Calculate all the
=
y_{j} ( 1  y_{j} )
Σ _{k}
(
w_{jk}
)

Calculate all the
=
I_{i}

For all i,j:
 Repeat (next exemplar).
See sample code.