{"id":"arxiv/2610.20266","url":"https://arxiv.org/abs/2610.20266","tags":["math.NA"],"score":0,"title":"Wavelet compression for sampling deep Gaussian processes","domain":"math.NA","source":"arxiv","authors":["Oskar Broukal","Michael Feischl","Conor Osborne"],"reviews":0,"subject":"math.NA","summary":"Deep Gaussian processes (DGPs) use iterative sampling to create non-stationary random fields. This is notoriously expensive because the covariance at each non-initial layer depends on the realised field from the layer below, so it changes with every sample. Moreover, standard fast sampling techniques like circulant embedding or matrix compression fail since the underlying kernel usually lacks the necessary regularity and structure. We develop and analyse a wavelet compression method for convolution Matérn DGPs. At each layer the covariance matrix is compressed using wavelets and a Lanczos…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-10-14","reactions":{}}