Lecture WS 25/26 Advanced Topics in Scientific Computing
Fast kernel methods
Kernels, understood as functions of two variables, arise in the sciences, statistics, and machine learning. While the nature of their appearance differs depending on the context where they arise, computational methods often have to perform rather similar operations. The fundamental problem for these computational methods is that kernel matrices are usually dense, implying at least quadratic and thus prohibitive computational complexity in storage and computation time at best. Within the lecture we will discuss various applications and computational methods and how to exploit structure in the kernels for faster computational methods.
eCampus: Link