Post date: Apr 07, 2014 9:32:9 AM
MaDaMS thesis student Johan von Tangen Sivertsen has recently completed his MSc thesis on kernel approximation, a tool for compactly summarizing high-dimensional data in a form useful for machine learning applications. In particular, the methods he has explored allows use of much larger training data sets within a certain time (e.g. SVMs), which could in turn lead to more precise predictions. His thesis and C++ source code is available at https://bitbucket.org/johanvts/fastkernel