search

dswah / sgcrfpy

SGCRFpy: Sparse Gaussian Conditional Random Fields in Python
MIT License
29 stars 5 forks source link

add max number of backtracking iterations, and early stopping #11

Open dswah opened 8 years ago

dswah commented 8 years ago

look up the stopping criterion that uses norm of gradient and norm of loss!

dswah commented 7 years ago

sometimes, the algorithm will propose a step direction, and then spend a long time backtracking until the learning rate is machine epsilon. perhaps instead of taking a really small step in a bad direction, we should just stop?