Neural Networks Toolbox BUG

As I asked a month ago, neural network structs can not be used in vector or matrix configurations. When it is stracted a neural network struct from a vector or matrix, it is not recognized as a network struct. This is a failure of Neural Networks Toolbox in versions 2011 and 2012 of Matlab.

I hope that Matlab developers will fix this bug as soon as possible.

Thanks in advance

[Walter Roberson]

What if you use cell arrays ?

[Ryan G]

You can report a bug directly so that the MathWorks team is aware of the issue.

[Akhan Khan]

I think I am also facing problem in neural network toolbox due to this bug. Using nntool I trained my network in R2011b but when I loaded the mat file in 2012a, network struct was not recognised and it gives me the warning " Could not determine the fields for class network by calling the constructor with no input arguments.The object net has been converted to a structure." I tried to create the network by "net=network;" but its not working in 2012a however it works in other versions. Please guide me if there is some solution for it.

[David Buldain]

Sorry Walter, I have more than 20 functions running in Matlab for practises in a university subject that were working perfectly with Matlab version 2010, so I have no intention to change that software that really works for something that possibly doesn´t...

[David Buldain]

Akhan, if your neural network struct was isolated (not in a vector or matrix), and only was recorded in a .mat file, you can solve the problem by placing this command before the load call: %#function network