Marginalizing stacked linear denoising autoencoders

Stacked denoising autoencoders (SDAs) have been successfully used to learn new representations for domain adaptation. They have attained record accuracy on standard benchmark tasks of sentiment ana...