This is another simple test of how deep-belief-network principles work (like this http://www.youtube.com/watch?v=8BqELwn0biA ), this time trained with the 600 first digits of the training set of the MNIST handwritten digit database and later refined with some back-propagation. The video shows how the network performs on the novel 150 digits of the MNIST test set. For implementation I used Python with Pylab and MDP's RBM nodes. Music by Madame Sufi.
Nice. Now translate the test image a few pixels and see what the result is.
rikio135 8 months ago
I don't get it. Whats the point?
Houshalter 1 year ago