Early Stopping. The Most Popular Regularization Technique In Machine Learning.

[@paulallen1597]

Excellent high-level explanation of this topic. 10/10. Thank you for your hard work!

[@underfitted]

Glad you liked it!

[@limotto9452]

You sounds like a hero of all Machine Learning realm. Thank you very much for the video sir.

[@emresdance]

You've got yourself a subscriber brother, great edit, clarity.

[@underfitted]

Thanks for the sub!

[@11aniketkumar]

Wow, your teaching skills are excellent

[@underfitted]

Thanks!

[@erfanelmtalab1615]

Supreme edit ! can't wait to see your channel grow ...

[@srishtigupta5507]

One of the most interesting video to learn from. Thanks!

[@nedafiroz514]

I want to say one thing, As many times I see your videos, i get inspired to work on the suggestion and improve my model. ❤The best explanation ever. Watching over several times

[@graysadler4085]

This is exactly what I’m building! I’m creating variable training sets to see which training set size has best performance.

[@underfitted]

Sounds great!

[@a7med7x7]

I just discovered one of the best channels

[@PritishMishra]

Awesome Explanation and The Thumbnail is lit 🔥

[@ЕгорАбросимов-л2о]

Great stuff! Looking forward to hearing more insights!

[@karlbooklover]

implemented early stopping auto-saving best model for os-cnn, great explanation! I love your content

[@underfitted]

Glad it was helpful!

[@whilstblower901]

Sir your Explanation is excellent please make Videos on Regularization for Deep learning Parameter norm Penalties, Norm Penalties as Constrained Optimization

[@alienm.nunezrivero2996]

That was awesome! My teacher, my mentor, you are just the person I had the opportunity to learn from the most in my whole carrier. Thank you for sharing stuff like this and also for the time we spent working together in the past. I noticed you are now doing what you love the more... just keep teaching

[@underfitted]

Yo! What’s up! Thanks for the comment! Love ya, man!

[@alwaleedalattas8384]

Really like the way you explained it, thanks a lot

[@underfitted]

Glad it was helpful!

[@abdullahyaser9073]

Fantastic explanation 👏

[@underfitted]

Glad you liked it

[@karimmerchaoui9736]

Just what I was looking for ! how about doing a full machine learning course and simplify concepts with the same approach you did in this video ?

[@underfitted]

Working on it.

[@VinayaWani]

great video! thanks.

[@M1911Original]

I wish you could provide a technical walkthrough alongside the theoretical

[@ThiagoSilvaOfficial]

Nice video, thanks

[@roshanaryal310]

Great explanation as always, Santiago 💯💯🔥

[@underfitted]

Appreciate it!

[@Naeem2460]

Amazingly creative explanation, new sub ❤

[@aayushpatil7514]

0:50 isn't this model over fitting to data?

[@HistoryUnlocked-fi3er]

Kind of…

[@kozaTG]

i hate how he tricks you into watching another video and you can't ignore because of how good the one you're currently watching is

[@varunahlawat9013]

OMG superbbbbbbbbbbbbb, today itself I came across this confusion (I've just started ML, today started with linear regression, so it's not always linear ahh satisfaction :))

[@underfitted]

Glad it was helpful!

[@dimasveliz6745]

nice analogy

[@underfitted]

Thanks :)

[@TheInevitableHulk]

You could tie the deceleration of the validation loss to the learn rate after a certain threshold.

[@underfitted]

Right on

[@thomaswijgerse723]

Early stopping is kinda frowned upon right? Since it does regularization and training at the same time. I think the preferred methods are L1, L2 and dropout iirc?

[@nicholasmarshall5775]

Great video, thanks! At 05:33, you mention the requirement of a "performance metric" - isn't this just the loss function the model is being optimized for?

[@underfitted]

It depends. Sometimes you need something business-specific. For example, what’s the impact of the model in real life?

[@pandabear_77]

Sir, I had a question. What would be difference between num of boost rounds vs early stopping rounds, since both are available as parameters in xgb.train. I am a little lost as to what would be difference between the two ? Any clarification on the same is appreciated.

[@stephenpaek9175]

Amazing content, please keep going Santiago

[@underfitted]

Thanks, will do!

[@shavilyarajput5477]

As far as i know santiago , epochs are in neural network right how can i use early stopping in ML ? I know it might be a noob question but i didn't really get it.

[@underfitted]

Any algorithm that relies on an iterative approach can benefit from early stopping. But yes, we primarily use it with neural networks.

[@tzvi7989]

Great video. Guess you're still limited by the input data though - especially in the omics fields

[@headshock1111]

Wait was this supposed to be an allegory for self improvement or what

[@vinaykamath6628]

Is it just me or does this guy sound like PewDiePie

[@StephenGillie]

Can an AI be trained to be better at training AI? You're a machine learning expert with a ton of manual work - why not get machine learning to do it for you?

[@MadgulkarAdityaManoj]

poor show