Platform

Platform

How to obtain optimal samples in Latin Hypercube Sampling in optiSlang?

    • agarwa20
      Subscriber

      What is the relation between the number of input design parameters and the number of samples required for Latin hypercube sampling.

    • Markus
      Ansys Employee

      Hi agarwa20,

      optiSLang automatically creates the sampling (optimally placing Designs in the Design Space) for you.
      Your question is probably, how many samples are required for a given set of input parameters?

      The required number of samples depends on the number of important input variables (the important variables can be seen in the Postprocessing using MOP and COP, more details are here. https://ansyshelp.ansys.com/account/secured?returnurl=/Views/Secured/corp/v222/en/opti_multi_disc/opti_multi_disc_metamodel.html).

      Using a too small number of samples in high dimensional problems may lead to a wrong selection or deselection of minor important or unimportant input variables.
      Best practices are: 
      - Use 2 times more samples than input variables (if possible) to reduce input correlation errors and to reduce probability of wrong selection or deselection.
      - Use at least 50 to 100 samples in order to get suitable estimates of variance-based input variable importance.

      Best regards,
      Markus

Viewing 1 reply thread
  • You must be logged in to reply to this topic.