Skip to main content

Table 5 Theoretical complexity per time step, where \(n\) represents the subspace dimension, \(p\) represents the parameter dimension, and \(N_\text {training}\) represents the number of training points

From: Non-intrusive nonlinear model reduction via machine learning approximations to low-dimensional operators

Method Complexity Comments
SVR 2 O(\((n+ p) nN_\text {training}\))  
SVR 3 O(\((n+ p) nN_\text {training}\))  
SVR rbf O(\((n+ p) nN_\text {training}\))  
Random Forest O(\(nN_\text {trees} N_\text {training}\log (N_\text {training})\)) \(N_\text {trees}\): number of decision trees
Boosting O(\(nN_\text {learners}\)) \(N_\text {learners}\): number of weak learners
kNN O(\((n+ p) nN_\text {training}+ K nN_\text {training}\)) K: number of nearest neighbors
VKOGA O(\((n+ p) N_\text {functions}\)) \(N_\text {functions}\): number of kernel functions
SINDy2 O(\(nN_\text {bases}\)) \(N_\text {bases}\): number of bases, \(N_\text {bases} < n^2\)