Kernel Methods

  • Describe a new method for performing kernel principal component analysis online and with a fast convergence rate. The method follows the Rayleigh quotient to obtain a fixed point update rule to extract the leading eigenvalue and eigenvector. Online deflation is used to estimate the remaining components. These operations are performed in reproducing kernel Hilbert space (RKHS) with linear order memory and computation complexity.
For more details go to Online kernel RQA page.


  • Implement content addressable memories (CAM) in a reproducing kernel Hilbert space (RKHS) to increase the amount of information stored. CAMs are one of the few technologies that provide the capability to store and retrieve information based on content. Even more useful is their ability to recall data from noisy or incomplete inputs. However, the input data dimension limits the amount of data that CAMs can store and successfully retrieve. We lifting this CAM limitation by implementing them in RKHS where the input dimension is practically infinite.
For more details go to Content Addressable Memories in RKHS page.



(more details coming soon)