We review a letter communicated by Yann Le Cun and published in Neural Computation in 2006, details a fast learning algorithm for deep belief networks. Authored by Geoffrey E. Hinton, Simon Osindero, and Yee-Whye Teh, the paper introduces "complementary priors" to simplify inference in complex belief nets. They propose a greedy, layer-by-layer learning approach that initializes a deep, directed belief network, followed by a fine-tuning procedure using a contrastive variant of the wake-sleep algorithm. The authors demonstrate the algorithm's effectiveness by achieving superior handwritten digit classification on the MNIST database, outperforming other leading discriminative methods. This work highlights the advantages of generative models for machine learning tasks.