Why can't Matlab R2010 load neural network objects from R2007?

Question

I have a neural network that I saved in a .mat file in Matlab 2007.

I am trying to load a neural network from a file using load filename.mat . This worked fine with the R2007b and R2008 versions, but when I try to load into R2010b, I get the following message:

A warning. An error occurred while starting the class loadobj method. The object loaded from the MAT file was a copy of the object before the loading method was started. The remaining variables were also loaded from the MAT file.
Error found:
Reference to a nonexistent field 'name'

The strange thing is that it seems that a variable carrying a neural network exists, but it does not work properly. None of the features using the neural network work in R2010.

Has anyone encountered a similar problem? How to keep Neural Network so that it is compatible with Matlab R2010?
Or even better, how can I load it correctly in Matlab 2010?

Answer 1

Yes , it seems that other people have encountered this problem (neural networks do not load in MATLAB 2010), according to this MATLAB Central , which suggests Andreas Goser (Mathworks Technical Support Manager):

If you see this problem in the future, contact technical support and refer to solution 1-EGNM3S.

Obviously, you need a "fixed net.m file" that technical support can provide. I have not tested it, since I myself do not have this problem.

Having carefully read the error message, I would say that the variable that you see loaded contains an "unprocessed" object that was saved in a .MAT file, quite possibly a struct. You can run whos myvarname to check. To “animate” this structure into an object, it must be handled by the overloaded loadobj function for its object type, which is usually automatically called by the MATLAB load function. In this case, loadobj , which comes with the neuron network object in R2010, does not seem to be fully compatible with the feedback ...

Answer 2

I suspect that the basic objects for the neural network have changed since 2007 and that the definition in R2010b is incompatible with the data stored in the MAT file.

I recommend answering Jonas. In addition, you can try to download it and save it in the intermediate MATLAB release or save data from neural network objects and rebuild them in R2010b.