G. Ferrari-Trecate, C.K.I Williams, and M. Opper. Finite-dimensional approximation of Gaussian processes. In M. Kearns, S. Solla, and D. Cohn, editors, Advances in Neural Information Processing Systems , volume 11, pages 218--224. MIT Press, 1999.
Gaussian process (GP) prediction suffers from O(n^3) scaling with the dataset size n. By using a finite-dimensional basis to approximate the GPpredictor, the computational complexity can be reduced. We derive optimalfinite-dimensional predictors under a number of assumptions, and show thesuperiority of these predictors over the Projected Bayes Regression method(which is asymptotically optimal). We also show how to calculate the minimalmodel size for a given n. The calculations are backed up by numericalexperiments.