“Measuring the sensitivity of Gaussian processes to kernel choice”

Rob Trangucci points us to this paper by William Stephenson, Soumya Ghosh, Tin Nguyen, Mikhail Yurochkin, Sameer Deshpande, and Tamara Broderick. I’m posting it here because it involves GPs, so Aki should be interested too.

Related ideas:

Static sensitivity analysis (for example section 6.3 here)

An automatic finite-sample robustness metric (by Broderick, Giordano, and Meager)

Covariances, Robustness, and Variational Bayes (by Giordano, Broderick, and Jordan)

Nothing new from me here, just the usual topic of trying to develop tools for understanding fitted models by considering various versions of d(inference)/d(input). I’m blogging it because it will be easier to find things that way.

Leave a Reply

Your email address will not be published. Required fields are marked *