Fig. 7
From: Richer fusion network for breast cancer classification based on multimodal data
Comparative analysis using different dimension raising methods: Copying features, Interactive and Polynomial features (IP), AutoEncoder features (AE), Denoising AutoEncoder features (DAE), and Denoising AutoEncoder with L1 regularization features (DAE-L1). Except for the different dimension raising methods, the other experimental settings were controlled as follows: Structured data (580D) + Pathological image (3, 4, 5 convolution layer, 1280D) + Fully connected layers (500, 100, 2)