Reconstructions from 30-dimensional codes
			• Top row is the data • Second row is the 30-D autoencoder • Third row is 30-D logistic PCA which works much better than standard PCA