I am using matlab neural network functions with feedforward network. The algorithm by default creates fully connected network, is it possible to create a network which is not fully connected and is it possible to delete some specific neuron connection? Just to clarify that i am using neural net for energy demand forecasting and ultimately it is a regression problem. Will appreciate if some one can help me with this.
Asked
Active
Viewed 112 times
0
-
1The algorithm by default chooses the best network. Sometimes, that network ma have zero weights in a neuron, thus making it disconnected. If the algorithm did not disconnect it, then its better the way it is now. Why would you want to discconect one? What is your purpose there? – Ander Biguri Feb 07 '18 at 18:40
-
Weel i am trying to prune the network, so the weights with very low value are made zero. However, after that i want to retrain the network without that neuron, this means the particular neuron wight should be deleted from the re-initialization of weights. – zafar Feb 09 '18 at 10:38
-
Then you may need to code your own NN. – Ander Biguri Feb 09 '18 at 10:39