fitcensemble settings to speed the process up

1 view (last 30 days)
I have a table with 260000 records, 9 fields of which 3 are categorical, the rest are double with the ninth being the target (0 or 1). I'm running fitcensemble as below :-
Mdl = fitcensemble(TestGB(:,1:8),TestGB(:,9),'OptimizeHyperparameters',...
'HyperparameterOptimizationOptions',struct('AcquisitionFunctionName','expected-improvement-plus'))
I've tried with less fields(3-4) and it runs reasonably quickly. With nine fields however it took overnight to do one round of calculations with 29 to go. As my machine only has 4 cores I thought I'd run it on an Amazon AWS compute VM with 16 Xeon processors only it didn't seem much quicker. Is there anything I'm doing wrong or that I could do to speed things up? Or am I just going to have to wait!
Stephen Gray

Accepted Answer

Aditya Patil
Aditya Patil on 3 Feb 2021
You can use the
struct('UseParallel',true)
name-value pair to improve performance of the hyperparameter optimization. This requires parallel computing toolbox.
See the extended capabilities section of the fitcensemble documentation for more information.

More Answers (0)

Community Treasure Hunt

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

Start Hunting!