BEAST2-Dev / nested-sampling

Nested sampling packagin for BEAST
GNU Lesser General Public License v2.1
11 stars 1 forks source link

Differences between MultiThreadedNS and NS when comparing Marginal Likelihood #12

Open dreyes17 opened 2 years ago

dreyes17 commented 2 years ago

Hello,

I am doing 10 serial executions with 6 particles using NS and 1 Multi threaded execution with 10 threads and 6 particles each thread using MultiThreadedNS.

The 10 ML values of both analysis are different, showing that the serial values are grouped at different points whereas the threaded values are grouped in a single point. For example in the serials executions some of them converged at approximately -4400, other -4600 and the other -4800. In the other hand taking a look at the 0.log ... N.log separetly they converge always very close to the same value, e.g -4400.

Do you have any clue why this is happening?

Thank you