Deep embedding kernel
Department
Information Technology
Document Type
Article
Publication Date
4-28-2019
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 predefined 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 and classification, and transfer learning, on different types of data including images, sequences, and regularly structured data.
Journal Title
Neurocomputing
Journal ISSN
0925-2312
Volume
339
First Page
292
Last Page
203
Digital Object Identifier (DOI)
10.1016/j.neucom.2019.02.037