The kernel Wiener Filter (kernel Dependency Estimation) in MATLAB.
Note: The algorithm dependes on the eigenvalue decomposition, thus only a few thousand of data samples for training dataset is applicable so far.
(In the near future, we will expand the algorithm for large scale data.)
There are 4 main functions:
1. kwiener_train -- Compute the kernel Wiener filter coefficient.
2. kwiener_predict -- Compute the pre-image using kernel Wiener filter.
3. kernelg -- Compute the kernel Gram matrix using Gaussian kernel.
4. demo_kwiener -- Run the USPS filtering problem using kernel Wiener filter.
Following references were related to the code:
1. J. Weston et.al. ``Kernel Dependency Estimation,'' NIPS 2003
2. M. Yamada and M. R. Azimi-Sadjadi, ``Kernel Wiener Filter using Canonical Correlation Analysis Framework,'' IEEE SSP'05, Bordeaux, France, July 17-20, 2005.
3. C. Cortes et.al. ``A General Regression Technique for Learning Transductions,'' ICML, Bonn, Germany, Aug 7-11.
4. M. Yamada and M. R. Azimi-Sadjadi, ``Kernel Wiener Filter with Distance Constraint,'' ICASSP, Toulouse, France, May 14-19, 2006
5. Zhe Chen et.al, ``Correlative Learning: A Basis for Brain and Adaptive Systems,'' Wiley, Oct. 2007 (Section 4.6) |
