Fig. 8
From: Sparse multi-output Gaussian processes for online medical time series prediction
The estimated population-level basis kernels and corresponding Bq matrices for 1003 patients with heart failure in MIMIC-III data set. We show the kernels estimated (a) without a sparse prior and (b) with a sparse prior (Q′=4). The sparsity of the Bq matrices is calculated as the percentage of nearly zero entries (i.e., values ≤10−3). The units for length scale or period are (d) for days and (h) for hours