My implementation of the Kernel Flows algorithm (for details see: Kernel Flows: from learning kernels from data into the abyss, Houman Owhadi, Gene Ryan Yoo at https://arxiv.org/abs/1808.04475).

I am currently am not maintaining this code so use at your own risk. The branch called 1.0 is the one I used for my MSc at Imperial. The branch called 2.0 contains many improvements, but is less complete and is based on what I have learned since.

Generally, the non-parametric version is less stable than the non-parametric. Both require some fine-tuning and are not always optimal. The parametric folder contains a notebook containing a tutorial.

If you have any questions, feel free to email me at mdarcy at caltech dot edu.