Resnext50_32_4d for imbalanced dataset to get better accuracy?

Can i use resnext50_32_4d for imbalanced dataset to get better accuracy?
is it better than resnet with less samples?,
someone used the resnext without data augmentation , imbalanced data but they got good auc score with simple cross validation technique.

I’m not aware of any model architectures, which would be beneficial for a classification use case using an imbalanced dataset. Usually, you would use a weighted loss, add over/undersampling or manipulate the training procedure in another way to tackle the issue of an imbalanced dataset.

As usually, I would recommend to try out different architectures for your use case, as I don’t think there is a simple answer which model would work best for an unknown task.

PS: please don’t tag certain people, as this might discourage others to answer, and you might tag someone like me, who cannot give a good answer to the question. :wink:

1 Like

@ptrblck Thank u so much yar for ur reply, Am working on skin cancer dataset.

Am new to tis forum, expected immediate answer from someone that point of time(running out of time) so mentioned u yar. :yum:

hey guys any idea about resnext? expecting answer from all of u too😋

dont mistaken me😉