Improving the Architecture of an Autoencoder for Dimension Reduction

Dimension reduction is used by scientists to deal with huge amount of high-dimensional data because of the "curse of dimensionality". There exist many methods of dimension reduction, such as principal components analysis (PCA), Locally Linear Embedding (LLE), Stochastic Neighbor Embedding (SNE), etc. Auto encoder is also applied for dimension reduction recently. It uses deep learning to train the network and has been applied in image reconstruction successfully. However, one important problem in auto encoder application is how to find the best architecture of the network. In this paper, we propose an improved architecture of the auto encoder for dimension reduction. The experimental results show the effectiveness of the proposed method.