An Empirical Evaluation of Deep Architectures on Problems with Many Factors of Variation

by Hugo Larochelle, Dumitru Erhan, Aaron Courville, James Bergstra and Yoshua Bengio

This is a paper about several experiments that verify the effectiveness of deep structure. One thing I see might be interesting is the autoassociator. It is a little different from Hinton's autoencoders. I am still considering how to add supervised information in the greedy training of DBNs. Once done, it might be useful for several problems. Hope I can figure it out ASAP.