"Marko Kolarek" <firstname.lastname@example.org> wrote in message <email@example.com>... > Hello everyone, > > I am looking for a neural network implementation that enables on-line learning. First I tried looking for a Cascade Correlated ANN implementation, but to no avail. I then discovered that Radial Basis Function ANNs could serve my purpose, but I'm still not sure how to implement the adding of new inputs in an already trained network. > > My only lead for now is this topic: http://www.mathworks.com/matlabcentral/newsreader/view_thread/122706 . > > Does anyone know has the situation improved since 2006.? Is it possible to add new inputs to an existing (trained) network? >
Unfortunately, newrb does not allow the specification of an initial network configuration.
My recommendation is to modify newrb with an option to begin with an initial specified configuration.
Then post the code or put it on the Link Exchange.