Machine Learning

We focus on the mathematical aspects of Learning Theory to the purpose of developing algorithms which can effectively learn the solution to a given problem from small samples. Our approach is based on the theory of regularization of ill-posed inverse problems and uses methods from functional analysis, convex analysis, and non-parametric statistics.
We are investigating:

  • Regularized kernel methods
  • Sparsity based regularization and variable selection
  • Manifold learning by spectral methods