Unfolding Kernel embeddings of graphs: Enhancing class separation through manifold learning
作者：
Highlights：
• We use manifold learning techniques to enhance the performance of graph kernels.
• We try to increase the class separation by unfolding the embedding manifold.
• We propose a procedure to find the optimal embedding of the data.
• The kernels which neglect the locational information show the largest improvement.
• The unfolding of the space helps to reduce the performance gap between the kernels.
摘要
Highlights•We use manifold learning techniques to enhance the performance of graph kernels.•We try to increase the class separation by unfolding the embedding manifold.•We propose a procedure to find the optimal embedding of the data.•The kernels which neglect the locational information show the largest improvement.•The unfolding of the space helps to reduce the performance gap between the kernels.
论文关键词：Kernel learning,Kernel unfolding,Graph kernels,Manifold learning
论文评审过程：Received 14 February 2014, Revised 19 March 2015, Accepted 20 March 2015, Available online 30 March 2015, Version of Record 16 July 2015.
论文官网地址：https://doi.org/10.1016/j.patcog.2015.03.018
