12. Performance

The code is optimized as far as my ability with python allows, and hopefully more improvements in this area will be made in the future.

Nonetheless, there are a number of simple things the user can do to reduce the time it takes ASteCA to process a cluster. Be aware that each one has a cost though, mainly in that the final result will lack estimation of certain parameters when some of these options are used.

1- The KDE p-value test can take up a substantial amount of time to finish. If the cluster is a clear physical entity (i.e.: there are no doubts about its existence as a true clustering of stars) then you can safely skip this test.

2- If you already ran the Bayesian decontamination algorithm function for a given cluster, and the center, radius and number of field regions remains the same in the params_input.dat file, then you can use the members file already generated to avoid running this function again. To do this, in the Bayesian decontamination algorithm block select the read mode, then cut the XXX_memb.dat file generated (where XXX is the name of the cluster; this file is saved along with the output images) and paste it in the input/ folder, next to the file that contains the photometric data of the cluster.

This way the code will not run this function again, instead it will read the membership probability values already generated.

Warning

This can only be done if the structural parameters (center and radius), the error rejection function used, and the number of field regions defined remains the same. Otherwise the values for the membership probabilities may change.

3- Lower the bootstrap number in the Best synthetic cluster fit function. This option must be used with care and avoided whenever possible. The bootstrap process assigns errors to the cluster’s fundamental parameters (metallicity, age, extinction, distance modulus, mass, binarity) found by the genetic algorithm. If you disable this feature (by setting a value of 1 or 0) two things will happen: first, parameters will have no errors estimation (which is not good); second, the processing time of the GA will be significantly reduced.

The reduced processing time is due to the bootstrap function being the one that takes up most of the running time in the code. This is so because, to obtain the errors estimates, it runs the GA again a number N of times. For example, if the GA takes 1 hour to complete and the bootstrap function was set to run N=10 times, it will take the bootstrap function approximately 10 hours to finish.

Important

One of the biggest advantages of using ASteCA is the possibility of associating statistically meaningful errors to the cluster’s parameters. A parameter that has an assigned value but no error estimate is substantially less valuable (and trustworthy) than one with a reasonable associated error.