I'm currently experimenting with the python bindings for FANN and the following article helped me along pretty well:
http://jansipke.nl/using-fann-with-python/
One thing that bothers me about this approach is, calling the train_on_file method results in a report being printed to stdout. Instead, I'd prefer to have something like a return value or exception handling, so my python code can be aware of the result. I'm considering embedding pyfann in a python program, but in order to achieve that I'd have to find a way to let it report back to python, not to me. I can't find any reference to this in the docs though. Any thoughts?
