Scalable Kernel Methods in Machine Learning

This repository contains my implementations made during the compact course "Scalable Kernel Methods in Machine Learning" by George Biros at Munich during June/July 2016.

All files can be run using Matlab. Parameters to change are always given in the first lines of the scripts. All notations and variable names are used as in the course and may differ from other implementations. The scripts are not optimized since learning the concepts is my main goal during this course. Feel free to use the code for learning purposes.

The repository is organized as follows:

• data: input data (.mat files)
• densityestimation: runnable Matlab files for estimating a Gaussian probability distribution from some samples
• functions: utility functions used within other files, in particular many different kernels
• kernel: some examples of kernel methods for the supervised classification problem
• material: slides and papers
• mnist: several scripts to classify hand-written digits