Bias introduced by GMM - Githubissues

GrainLearning / grainLearning

A Bayesian uncertainty quantification toolbox for discrete and continuum numerical models of granular materials, developed by various projects of the University of Twente (NL), the Netherlands eScience Center (NL), University of Newcastle (AU), and Hiroshima University (JP).

https://grainlearning.readthedocs.io/

GNU General Public License v2.0

9 stars 1 forks source link

Bias introduced by GMM #15

Open chyalexcheng opened 1 year ago

chyalexcheng commented 1 year ago

GMM introduces bias to the sampler. Two ways to remedy this.

slice sampling

train gmm
define the lowest score from the trained gmm on expanded samples as the threshold
generate samples from a low-discrepancy sequence (which one has the property of not being affected by cutting or truncating?)
include those higher than the threshold

restore the uniform prior before the data assimilation loop

chyalexcheng commented 1 year ago

Slice sampling now works. However, old "low-discrepancy" parameter samples are still sampled by Halton. Need to copy old simulations to new iterations.