OpenKernel Library

OpenKernel is a library for creating, combining and using kernels for machine learning applications. The current focus of library is on rational kernels. It is based on the OpenFst library.

This library is being developed by C. Allauzen, M. Mohri and A. Rostamizadeh. It is intended to be comprehensive, flexible, efficient and scale well to large-scale problems. It is an open source project distributed under the Apache license. This work has been partially supported by Google Inc.

New version available now, includes algorithms for learning kernels and kernel ridge regression.

• Background Material
• Creating and using rational kernels
  • Quick Tour
    • Using the library
      • Preparing your data
      • Creating an n -gram kernel
      • Computing the kernel matrix
    • Using the LIBSVM plugin
  • A detailed example: a subset of Reuters-21578
• Learning kernels
  • Quick Tour
  • Full documentation
  • Table of implemented features
  • Example of datasets and results
• Kernel ridge regression
• Download