Graceful ways for stuck Monte Carlo instances to fail and resample
Hey everyone, (ahem... mostly talking to you, Stefan!)
I'm trying to run a set of Monte Carlo simulations and there are a subset of cases that get stuck in a seemingly unending loop of minuscule time steps. Is there a graceful way to let iTOUGH know these are a lost cause? For example, if I set MCYC to a hard limit (i.e., a positive number less than 9999), and it reaches the "end" of the simulation without collecting all of the outputs requested in the "> OBSERVATIONS" level sub commands, will it recognize this run was not successful and try resampling parameters?
I admit that I did not do a thorough review of the (excellently written!) 130 page iTOUGH2 manual, which may contain an answer already. I wasn't able to find it with my lazy word-search approach. : )
Depending on the iTOUGH2 version you have, you can indeed perform CPR on a failed forward run within Monte Carlo simulations (or most other application modes) by humming Bee Gees "Staying Alive". So add the command:
> computation >> stop >>> stay alive (: max_number_of_failures_before_calling_the_hearse)
As you suggested, set MCYC to the highest number you can tolerate (or any other convergence-failure-triggering parameter) and add the "stay alive" command, and it should resample a failed Monte Carlo realization. (Disclaimer: I'm not sure how well it works in parallel).