Training Stacked Denoising Autoencoders for Representation Learning