CMStatistics 2022: Start Registration
View Submission - CMStatistics
B1798
Title: Feature representation learning in neural networks guided by the sliced inverse regression Authors:  Su-Yun Huang - Academia Sinica (Taiwan) [presenting]
Yan-Bin Chen - Academia Sinica (Taiwan)
Abstract: Due to demands from recognition of digital biomedical images, a lot of interest is being paid to finding feature representations in neural networks. We will share a neural network-based feature representation learning method guided by the notion of Sliced Inverse Regression (SIR). The proposed method works on deep neural networks with the SIR as a neural network layer. The proposed SIR layer is expected to improve image classification accuracy. We will present some numerical examples to demonstrate the usage of SIR-guided feature representation learning.