"Kernelized" Self-Organizing Maps for Structured Data

The suitability of the well known kernels for trees, and the l esser known Self- Organizing Map for Structures for categorization tasks on structured data is inves- tigated in this paper. It is shown that a suitable combinatio n of the two approaches, by defining new kernels on the activation map of a Self-Organizing Map for Structures, can result in a system that is significantly more accur ate for categorization tasks on structured data. The effectiveness of the proposed approach is demon- strated experimentally on a relatively large corpus of XML formatted data.