There is a newer version of the record available.

Published March 2, 2021 | Version v0.1.19
Software Open

refnx/refnx: REL: v0.1.19

Description

  • MCMC sampling in GUI now plots parameter value vs step number, so you can see how many samples to burn
  • Fixed GUI crash when trying to make a snapshot with a name that has already been used.
  • Give BaseObjective a default weighted attribute, enabling it to be used in GlobalObjective.
  • Fallback to numerical estimation of Hessian/Covariance matrix if an Objective cannot calculate a residuals array (e.g. mixed use of BaseObjective/Objective in a GlobalObjective).
  • Vendor scipy.stats.qmc.
  • Use Latin Hyper Cube sampling to initialise MCMC walkers when initialising with prior. This ensures a good distribution of walkers and no clustering.

Files

refnx/refnx-v0.1.19.zip

Files (11.1 MB)

Name Size Download all
md5:b7c2f6e9d8bfb2af22067c33918427e1
11.1 MB Preview Download

Additional details

Related works