Stopping criterion broken about percolator

[MatthewThe]

The early stopping criterion is only checked if you set the --quick-validation parameter, otherwise it will indeed run all 10 iterations.

[wsnoble]

Why? I can't imagine any scenario where you want to keep training after convergence. Is the idea that sometimes it appears to converge and then later finds a significant additional number of PSMs?

[wsnoble]

At any rate, if this is the behavior, then the parameter "maxiter" should be renamed to "numiter." The documentation is currently incorrect.

[kevinkovalchik]

The early stopping criterion is only checked if you set the --quick-validation parameter, otherwise it will indeed run all 10 iterations.

I know this is an old thread, but the documentation regarding this feature really does not make any sense. From the documentation:

-i
--maxiter Maximal number of iterations. Default = 10.
-x
--quick-validation | Quicker execution by reduced internal cross-validation.

There is no way to know that --quick-validation enables early stopping. Perhaps the addition of an --early-stopping flag would help.