A Novel Scatter-enhanced Correlation Feature Learning Method
Shuzhi Su, Jun Feng Xie, Yanmin Zhu, Xingzhu Liang · 2021
Canonical Correlation Analysis (CCA) is an essential algorithm in the feature learning field. However, it does not utilize supervised information, and it failed to solve nonlinear problems. Therefore, this paper proposes a novel feature learning algorithm called Scatter-enhanced Canonical Correlation Analysis (SeCCA). This paper integrates the internal structure information and supervised information of the data and embeds them into the canonical correlation framework. The excellent image recognition performance of this algorithm can be demonstrated by extensive experimental results.