Gaussian process regression optimization

3 views (last 30 days)
H R
H R on 4 Sep 2021
Commented: H R on 26 Jul 2022
Can anyone please let me know how can I exclude some Kernel functions from being optimized in the following regression process?
In other words, I would like to only include 'matern32', 'matern52', and 'ardmatern52' in the optimization process to fit a Gaussian regression model to the data. When I use the following code lots of other unwanted kernel functions are also tested.
thisModel = fitrgp(X,Y,'ConstantSigma' ,true, 'Sigma',0.001,'BasisFunction','none',...
'OptimizeHyperparameters',{'KernelFunction','KernelScale','Standardize'}','HyperparameterOptimizationOptions',...
struct('Optimizer', 'bayesopt', 'MaxObjectiveEvaluations', 45,'AcquisitionFunctionName','expected-improvement-per-second-plus','ShowPlots', false, 'UseParallel', true));
Many thanks.
  2 Comments
Hiro Yoshino
Hiro Yoshino on 26 Jul 2022
Seems like there are no ways to have subset of all the Kernels available.
how about running fitrgp(s) independently with the kernels fixed? You could use Parallel option too.
H R
H R on 26 Jul 2022
I actually found a way to do this. You basically need to define the optimizableVariable separatly and then feed it to OptimizeHyperparameters.

Sign in to comment.

Answers (0)

Tags

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!