Dimensionality reduction strategy based on auto-encoder
Yasi Wang, Hongxun Yao, Sicheng Zhao, Ying Zheng · 2015
Auto-encoder is a tricky three-layered neural network, which constructs the "building block" of deep learning that has been demonstrated to achieve good performance in various domains. In this paper, we focus on auto-encoder's dimensionality reduction ability, and try to investigate whether auto-encoder has some kind of good property that can accumulate when being stacked, thus contribute to the success of deep learning.