Recent research has been devoted to learning algorithms for deep architectures such as Deep Belief Networks and stacks of auto-encoder variants, with impressive results obtained in several areas. The best results obtained on supervised learning tasks involve an unsupervised learning component, usually in an unsupervised pre-training phase, with a generative model. Even though these new algorithms have enabled training deep models fine-tuned with a discriminant criterion, many questions remain as to the nature of this difficult learning problem. The main question investigated here is the following: why does unsupervised pre-training work and why does it work so well? Answering these questions is important if learning in deep architectures is to be further improved. We propose several explanatory hypotheses and test them through extensive simulations. We empirically show the influence of unsupervised pre-training with respect to architecture depth, model capacity, and number of training examples. The experiments confirm and clarify the advantage of unsupervised pre-training. The results suggest that unsupervised pre-training guides the learning towards basins of attraction of minima that are better in terms of the underlying data distribution; the evidence from these results supports an unusual regularization explanation for the effect of pre-training.