moineau can be executed locally to help the developper compare the
results from its IP and the moineau results, at every state of the
program. The two implementation are behaving the same way.
Respect the following steps to compile and launch moineau:
make framesto generate theframes.cfile from MNIST datamake parse NEURONS=xto generate thenet.cfile inmoineau/src,xmust be100or200(100by default) to select the number of neurons of the first layermake NEURONS=x,xmust be the same as the firstmake parsecommand (100by default)./moineauto execute
You can also generate less than the 1000 frames and use another MNIST data
file. Check in extra/frames.py for options.