feeding training and testing data into network

Hello,

I hope you can help me :)

I am trying to train a radial basis neural network with a 'training set', and test it against the 'testing set'. I do not want to select my 'testing set' randomly from a larger set. I have carefully selected my training/testing sets.

I am trying to train many networks in an attempt to find the best possible network for generalization by dynamically adjusting the free parameters.

I must use an radial basis neural network.

Note: I have found things like this net.divideParam.trainRatio = 70/100; % Adjust as desired net.divideParam.valRatio = 15/100; % Adjust as desired net.divideParam.testRatio = 15/100; % Adjust as desired but believe them to be too simplified.

Answer 1

[Greg Heath]

Those functions are not valid for the radial basis functions Just train with the training set. Then, when it is finished, test with the test set.

Unfortunately the RBFNN functions newrb and newrbe are neglected step-children.

They have few of the options given to the MLP functions (fitnet, patternnet, etc).

In particular,

 1. You cannot initialize weights before the first training loop.

 2. You cannot use validation and test sets.

 3. There is no training history output tr.

 4. ...

Thank you for formally accepting my answer

Greg