Hello,
I'm developing a communication protocol to secure data transfer over a tcp network. I want to use neural networks to exchange the keys and found a protocol specification in wikipedia:
http://en.wikipedia.org/wiki/Neural_cryptography#Tree_parity_machine[
^] and an implementation in Delphi:
Neural Cryptography[
^]
and recently a new project in F#:
http://neuralcryptography.codeplex.com/[
^]
I've tried to implement it in C# using aforge, but I think it does not have some necessary features to implement a neural cryptography protocol.
My problem is that I don't know how to design a neural network like the one in the articles!
I have used aforge but i don't know how to train the network beacuse I didn't found any of the learning algurithm below :
* Hebbian learning rule:
w_i^+=w_i+\sigma_ix_i\Theta(\sigma_i\tau)\Theta(\tau^A\tau^B)
* Anti-Hebbian learning rule:
w_i^+=w_i-\sigma_ix_i\Theta(\sigma_i\tau)\Theta(\tau^A\tau^B)
* Random walk:
w_i^+=w_i+x_i\Theta(\sigma_i\tau)\Theta(\tau^A\tau^B)
and how to define the Signum function (it seems that it does not support custom function) and how to control the out put of the hidden neurons!
I have tested all of the features of the aforge
Does aforge support such features?
Do you know any alternative library ?
Thanks in advance!