Fine Tuning vs Transfer Learning

preview_player
Показать описание
Here you go, we can understand the difference between the fine-tuning and transfer learning clearly here.
Рекомендации по теме
Комментарии
Автор

this is a great video. Most of the other stuff that came up when i searched for this was just literal PyTorch or tensorflow tutorials with very little to learn in terms of pure theory. I feel like this is super useful for people who already know how to use those tools and want to have a deeper understanding on this

himacri
Автор

Extremely simple and easy to understand, and covers the use case for both methods.
That helped me, thanks!!

bftlsqc
Автор

Loved the way u explained this concept. Thank You for this!

Ajay
Автор

Best explanation ever! Thank you so much!

Tech-rqcq
Автор

This is cool, always had this doubt in understanding... Thanks a lot..Your video made it clear..

sp
Автор

Happy new year!
Good and informative 2021...

sinikishan
Автор

Might it be worth mentioning that with fine-tuning, it might be advisable to use a smaller learning rate, so as not to clobber the original weights too much?

JuanUys
Автор

I have one doubt
In transfer learning also sometimes we will unfreeze last few layers for better performance. So will that be called fine tuning or transfer learning ?

SaiKiranAdusumilli
Автор

Thanks in advance it makes me clear my doubt, Sir

mayphyu
Автор

Can someone give me a hint over the weights please?
In transfer learning, we freeze the feature extraction layers so they keep their weights with no change, we retrain the classification layers. So does it mean that the classification layers have their weights from before and we retrain them meaning that we want to slightly change the weights of classification layers?

rosacanina
Автор

If I understand correctly, Transfer Learning is a proper subset of Fine Tuning.

duhfher
Автор

Very nice tutorial. How about the term "pre-training" (not sure if I made this up in my mind), or is it better to avoid this term to avoid confusion?

runerealm