adaptwb
Adapt network with weight and bias learning rules
Syntax
[net,ar,Ac] = adapt(net,Pd,T,Ai)
Description
This function is normally not called directly, but instead called indirectly through the
function adapt
after setting a network’s adaption function
(net.adaptFcn
) to this function.
[net,ar,Ac] = adapt(net,Pd,T,Ai)
takes these arguments,
net | Neural network |
Pd | Delayed processed input states and inputs |
T | Targets |
Ai | Initial layer delay states |
and returns
net | Neural network after adaption |
ar | Adaption record |
Ac | Combined initial layer states and layer outputs |
Examples
Linear layers use this adaption function. Here a linear layer with input delays of 0 and 1,
and a learning rate of 0.5, is created and adapted to produce some target data
t
when given some input data x
. The response is then
plotted, showing the network’s error going down over time.
x = {-1 0 1 0 1 1 -1 0 -1 1 0 1}; t = {-1 -1 1 1 1 2 0 -1 -1 0 1 1}; net = linearlayer([0 1],0.5); net.adaptFcn [net,y,e,xf] = adapt(net,x,t); plotresponse(t,y)
Version History
Introduced in R2010b