SSSK – Sparse Spatial Sample Kernel


Sparse spatial sample kernels (SSSK) are described in the following publications:


  • Kuksa, P., Huang, P. & Pavlovic, V. (2008), “Fast Protein Homology and Fold Detection with Sparse Spatial Sample Kernels”, In Int’l Conf. Pattern Recognition. Tampa, FL. Dec. 2008.
  • Kuksa, P., Huang, P. & Pavlovic, V. (2008), “Scalable Algorithms for String Kernels with Inexact Matching”, In Neural Information Processing Systems (NIPS). Vancouver, Canada. Dec. 2008.

For more details refer to our Spatial Kernels project page.


You can download two implementations of SSSK: a Matalb/cmex version and a standalone C version. Both versions produce SSSK Gram matrices, which can then be used for classification or clustering of sequences. The distributions include a README file, a sample file with sequences, and the corresponding reference outputs (kernel matrices).