Tree-structured multi-stage principal component analysis (TMPCA): Theory and applications

作者：

Highlights：

• An efficient dimension reduction method for sequence embedding.

• The proposal maintains high mutual information between its input and output

• Mathematical analysis of the proposal’s complexity, linearity, and orthonormality.

• Superior performance as compared to the state-of-the-art neural-network-based approaches

摘要

•An efficient dimension reduction method for sequence embedding.•The proposal maintains high mutual information between its input and output•Mathematical analysis of the proposal’s complexity, linearity, and orthonormality.•Superior performance as compared to the state-of-the-art neural-network-based approaches

论文关键词：Dimension reduction,Principal component analysis,Mutual information,Text classification,Embedding,Neural networks

论文评审过程：Received 28 July 2018, Revised 7 October 2018, Accepted 11 October 2018, Available online 11 October 2018, Version of Record 16 October 2018.

论文官网地址：https://doi.org/10.1016/j.eswa.2018.10.020