the bandwidth gamma and the decay of r_k - some experiments on Boston data

[el-hult]

Today on the seminar we had a little discussion about how choosing the tuning parameter gamma in RFF possibly could correspond to getting the appropriate decay in the spectum of the feature covariance.

I did some runs on that with the boston housing data. I hope it may be of use to someone. It is a notebook exported as a pdf.

boston_rank.pdf