Cancel some of the weights in a neural network

Hi,

Im using the network net = cascadeforwardnet() with two hidden layers. I would like to have connections from the input to the two hidden layers but not to the output layer. However cascadeforwardnet() automatically set these weights.

1. Can I cancel it ? ( i don't mean only to set the weight to zero)

2. In general, can i cancel some of the weights? for example from the i'th unit of the first hidden unit to the j'th unit in the second hidden layer ?

Answer 1

[Greg Heath]

I' pretty sure that if you check the custom network section in the documentation, you can learn how to set the learning rate of specified weights to zero.

Hope this helps.

Greg