Extracting and composing robust features with denoising autoencoders — latest news, discussion & facts — IJR