Neural Network for Random Sequence Generation

Hi,

I have designed a custom neural network for random sequence generation. I have my custom transfer function and I want to update the transfer function parameters on each cycle using a neurons own transformed output. Also since I'm not training the network with a data set, I want the outputs of the output layer to be xored and fed back as input to input layer neurons for next operation cycle. Please suggest how I can achieve these functions

[Thorsten]

What is the format of your input and output? How many, and what values can each have?

[Guru]

Hi,

Thanks for the reply. Right now I plan to have two inputs, 2 hidden layers with 6 neurons each and four outputs. For each input I'm planning to give as a 1-by-1 matrix with value inthe range of 0-1. The output also should be a 1-by-1 matrix in the range 0-1.

Answer 1

[Thorsten]

One way to do it would be to xor output 1 and 2 to get input 1, and output 3 and 4 to get input 2:

output = [ 1 0  1 1]; % sample output
new_input = [ xor(output(1), output(2)) xor(output(3), output(4))];

[Guru]

Thanks for the reply. How can I specify the new_input to be used by first layer.

[Thorsten]

input = new_input;