11-20-2016 08:31 PM
Is there a way to reduce the run time of the SAS gbm node in EM ? We are running HP forests and R inside the Open Source Integration node with comparable run times on the same data.
As soon as we run the SAS GBM with the same parameter sets as the R based GBM in the Open Source integration node the run time of the EM GBM nodes is 5-10 times slower.
Are there any system / node parameters that we can look at to speed up the processing.
11-21-2016 12:31 PM
11-21-2016 04:17 PM
Thanks for that. I am aware that GBM is still a serial procedure. I am also aware that HPFOREST runs in parallel even in SMP mode.
Is there a way to speed up the existing GBM node ? We are using GBM in R as our 'go-to' algorithm for most of our predictive models and we want to replicate that in SAS.
HOwever the SAS GBM is orders of magnitude slower than the R equivalent on exactly the same data.
SAS Viya is not on our horizon soon so we'll have to make do with SAS or R.
11-21-2016 05:50 PM