Improving the Architecture of an Autoencoder for Dimension Reduction | IEEE Conference Publication | IEEE Xplore