Using CNN, why doesn't my loss function go down? the stability is frightening.

after several iterations, the loss function remains unchanged, and the learning rate is still changed. I hope the great god will guide you what the problem is
this is the code https://github.com/yaoqinchua.
uses multi-perspective CNN (mvcnn)

.

Sorry is not the answer, because I also encountered the same situation

I don't know if the landlord has solved it


I didn't solve it either. Despair


first of all, understand that Loss does not necessarily shrink to zero!
it is recommended to check the loss graph. If the overall train-loss curve remains down or unchanged, as long as val-loss goes down or does not rise, there is nothing to worry about.
of course, at this time, you should also pay attention to train-acc and val-acc to see if the accuracy of the training set is declining

.
MySQL Query : SELECT * FROM `codeshelper`.`v9_news` WHERE status=99 AND catid='6' ORDER BY rand() LIMIT 5
MySQL Error : Disk full (/tmp/#sql-temptable-64f5-1b3610d-2bfef.MAI); waiting for someone to free some space... (errno: 28 "No space left on device")
MySQL Errno : 1021
Message : Disk full (/tmp/#sql-temptable-64f5-1b3610d-2bfef.MAI); waiting for someone to free some space... (errno: 28 "No space left on device")
Need Help?