Deep Embedding Kernel

TL;DR

The paper introduces Deep Embedding Kernel (DEK), a novel supervised learning approach that combines deep learning and kernel methods, enabling more adaptable and generalizable data representations for various tasks.

Contribution

DEK is a learnable kernel with a new deep architecture that improves data mapping and generalization over traditional kernels and deep learning models.

Findings

DEK outperforms traditional methods in identity detection and classification.

DEK demonstrates superior results in regression and dimension reduction.

DEK shows enhanced transfer learning capabilities.

Abstract

In this paper, we propose a novel supervised learning method that is called Deep Embedding Kernel (DEK). DEK combines the advantages of deep learning and kernel methods in a unified framework. More specifically, DEK is a learnable kernel represented by a newly designed deep architecture. Compared with pre-defined kernels, this kernel can be explicitly trained to map data to an optimized high-level feature space where data may have favorable features toward the application. Compared with typical deep learning using SoftMax or logistic regression as the top layer, DEK is expected to be more generalizable to new data. Experimental results show that DEK has superior performance than typical machine learning methods in identity detection, classification, regression, dimension reduction, and transfer learning.