In plenty of cases, fitting leads to state when R^2 in case of least squares method or likelihood in MCMC doesn't show any further progress for long time. Usually it fall into local minimum and won't change this state. We should add feature to interrupt such state prematurely.
Example (it stays in R^2 ~ 0.994 for more than 100 runs):
mikecokina
commented
4 years ago
In plenty of cases, fitting leads to state when R^2 in case of least squares method or likelihood in MCMC doesn't show any further progress for long time. Usually it fall into local minimum and won't change this state. We should add feature to interrupt such state prematurely.
Example (it stays in R^2 ~ 0.994 for more than 100 runs):