A DBN is equivalently just stacked RBMs. Most folks just use stacked autoencoders for the exact same reasons you'd use a DBN. Most "off the shelf" work people

A DBN is equivalently just stacked RBMs. Most folks just use stacked autoencoders for the exact same reasons you'd use a DBN. Most "off the shelf" work people do is fairly simple transfer learning on an imagenet cnn. That or if they do generative, most of the hype is focused on GANs now.