Kernel Methods
ABOUT
Kernel methods is a class of algorithms frequently used in the field of machine learning. Kernels can be used to analyse, process and compare many types of data. Since kernel methods use square matrices to process data, they can ananlyze images, molecules or even sequences. Moreover, Kernels have several properties. They can be inner products, measures of similarity and measures of funciton regularity. Most kernel methods, such as the SVM (Support Vector Machine), are based on two main concepts: The kernel trick which makes it possible to transform a linear method to a non-linear kernel method by replacing the dot product by a more general kernel and the representer theorem which is applied when the kernel is thought as regularisation operator.
RESOURCES
Vert, J. P., Tsuda, K., & Schölkopf, B. (2004). A primer on kernel methods. Kernel Methods in Computational Biology, 35-70.
http://www.kernel-methods.net
Shawe-Taylor, J.; Cristianini, N. (2004). Kernel Methods for Pattern Analysis. Cambridge University Press.
Liu, W.; Principe, J.; Haykin, S. (2010). Kernel Adaptive Filtering: A Comprehensive Introduction. Wiley.
