Similar loss, different results

by Rizhiy   Last Updated October 09, 2019 17:19 PM

I have trained multiple CNNs for image classification. I suspect there is something wrong with my training pipeline, since many of my experiments get very similar training loss at the end of training, but perform very differently on the test set.

What can be the cause of that?



Answers 1


since machine learning consists of stochastic methods unless you will set specific random seed on beginning, your model will be initialized with different values and this will lead to different values of model after training

quester
quester
October 09, 2019 17:01 PM

Related Questions


Updated October 19, 2017 12:19 PM

Updated May 16, 2017 22:19 PM

Updated July 04, 2018 12:19 PM

Updated October 06, 2018 15:19 PM