search

stephenslab / mixsqp

R package for fast maximum-likelihood estimation of mixture proportions using SQP.
https://bit.ly/2NtYHWT
Other
11 stars 7 forks source link

Run one or more EM updates in mix-SQP "outer loop" for more reliable convergence #38

Closed pcarbo closed 4 years ago

pcarbo commented 4 years ago

As suggested by @stephens999: "I’m still somewhat inclined to investigate the idea of doing one EM step per mixSQP outer loop—it just seems an attractive and cheap way to avoid some outlying problems."

pcarbo commented 4 years ago

Resolved as of commit 0d201ff.