Gradient Boost Machine Learning|How Gradient boost work in Machine Learning

[@prasaddalvi3017]

By far the best theoretical explanation on Gradient Boosting. Now I am very much clear on how Gradient Boosting works. Thank you very much for this detailed explanation

[@UnfoldDataScience]

Thanks Prasad.

[@arjunp3574]

This is the most simplified explanation of gradient boosting I've come across. Thank you, sir.

[@UnfoldDataScience]

Glad it was helpful Arjun.

[@denial4958]

Thank you sir it's the day before my exam and this concept was very unclear to me no matter how much I researched. Simply a life saver👏👏

[@shashankbajpai5659]

by far the simplest and the best explanation i could have for adaboost

[@UnfoldDataScience]

Thanks Shashank. Happy Learning. Stay Safe. tc

[@confused4596]

This channel is my Bible!! Thank you for creating ML content, Aman Sir

[@UnfoldDataScience]

Your comments are precious.

[@sriramapriyar6745]

I have no words to thank you for teaching this complex concept in a simple and effective way. My heartfelt thanks and keep going with the same spirit.

[@UnfoldDataScience]

Hello Sri, Thanks for you words. These are my motivation to create more content. Happy Learning. Tc :)

[@warpathcucucu]

mate, that's literally the best explanation for this topic on youtube

[@UnfoldDataScience]

Wow, thanks a lot.

[@Atlas-ck9vm]

Probably the greatest explanation of gradient boosting on the internet.

[@UnfoldDataScience]

Thanks a lot.

[@hvs147]

This is by far the clearest/best explanation on Gradient Boosting. Thanks man. God bless!

[@UnfoldDataScience]

Thanks Harsha. Happy Learning. Tc

[@preranatiwary7690]

Gradient boost is clear now! Thanks.

[@UnfoldDataScience]

Thank you Prerana:)

[@gg123das]

Best Gradient Boosting video on TH-cam!!!!

[@UnfoldDataScience]

Glad it was helpful Jafar. Happy Learning. Tc

[@dd1278]

Legend you are for explaining this so simply.

[@UnfoldDataScience]

Thanks again deb.

[@praveenkuthuru7439]

Great work !!! Really helps a common person to learn about the GB Algorithm in action in simple terms....Keep up your good work !!!!

[@UnfoldDataScience]

Thanks Praveen, will do!

[@OhSohomemade]

Hey Aman ..very well explained ... I am beginner and was looking for a easy and practical way of learning these concepts and you made it easy ..thanks very much ..appreciate the good work ..cheers

[@UnfoldDataScience]

It's my pleasure. Keep Learning. Stay Safe. tc :)

[@divyosmiley9]

Thank You, Sir. I read many papers but was so confused, but you made it clear.

[@UnfoldDataScience]

Thanks for your valuable comment.

[@naqiuddinadnan9156]

I need to study this by myself but mostly the explaination are not soo clear but u give great explaination 👍🏼👍🏼👍🏼

[@UnfoldDataScience]

Thanks Naqiuddin :)

[@sandeepnayak2427]

Its excellent, very much clearly step by step explained , Highly Appreciable ...You are Awesome ..

[@UnfoldDataScience]

Thanks Sandeep. Keep larning. Stay Safe !

[@divyanshumnit]

Thank you so much, Sir. I have watched it so many places but the clarity I got from your video. just watching this video I subscribed to your channel.

[@UnfoldDataScience]

Thanks Divyanshu.

[@GopiKumar-ny3xx]

Nice presentation.... Useful information

[@UnfoldDataScience]

Thanks a lot :)

[@adithyaboyapati]

Explanation is crisp and very clear.

[@UnfoldDataScience]

Thanks Adithya.

[@kayodeoyeniran2862]

Thank you for demystifying such a confusing concept. This is the best explanation by far!!!

[@UnfoldDataScience]

Thank you

[@amalaj4988]

Learning a lot from you sir! Crisp and clear points as usual :)

[@UnfoldDataScience]

Thanks Amol. Your comments motivate me to create more content 😊

[@RanjitSingh-rq1qx]

Super explanation with in less time. With mathmatics intuition. Tnq u sir for made this mind-blowing video ❤️🥰😊

[@ChengchengCai-v9k]

Thank you! Best explanation! I can understand it!

[@amithnambiar9818]

Thank you ! Never seen a video so detailed yet understandable about Gradient Boosting

[@UnfoldDataScience]

Thanks Amith. Happy Learning. Tc

[@IRFANSAMS]

@Unfold Data Science, Sir the way you explain complex topics in a simple manner is extraordinary

[@praneethaluru2601]

Very goood and elegant explanation of GBoost than others on TH-cam Sir...

[@UnfoldDataScience]

Thanks For watching Praneeth.

[@sudhavenugopal3726]

Well explained, where a beginner can understand this, thank you so much

[@UnfoldDataScience]

Thanks a lot Sudha :)

[@FarhanAhmed-xq3zx]

Very very simple and clear explanation.really awesome👌👌

[@UnfoldDataScience]

Thanks Farhan.

[@josephmart7528]

You have made my day with this Ensemble Explanations

[@UnfoldDataScience]

Thanks Joseph.

[@cedwin4]

Simple and best. Speechless! Thanks a lot :)

[@vivekbhatia8230]

Very nicely explained sir.. as u said it was not very clear in net.. after your explanation i can understand the working of the gradient boost model.

[@abiramimuthu6199]

I was running around so many videos for Gradient boosting.........Than you so much for your detailed explanation.....How does it work for a classsification problem?

[@UnfoldDataScience]

Hi Abirami, thank you for the feedback. It's difficult to explain the classification problem through comment. I ll probably create a video for the same :)

[@shubhankarde4732]

please create one video for classification as well.....

[@dorgeswati]

you are awesome . video shows the depth you have in understanding these algorithms well. keep it up

[@UnfoldDataScience]

Thanks a lot!

[@bangarrajumuppidu8354]

superb explanatio fanstastic!!

[@UnfoldDataScience]

Thanks Bangarraju.

[@hashir3719]

It's crystal clear mahn..! thank you

[@snehasivakumar9542]

Easy to understand. 😊👍

[@UnfoldDataScience]

Thank you Sneha :)

[@shashireddy7371]

Thanks for sharing your knowledge with great explanation .

[@UnfoldDataScience]

Welcome shashi, keep learning :)

[@KenzaLamnabhi-f8l]

Thank you for this video! really amazed by how you siplify complex concepts !
Keep them going please!

[@IRFANSAMS]

Aman sir, Allah will give you more success in your life

[@UnfoldDataScience]

Thanks Imran, your comment mean a lot to me.

[@shreyaanand6323]

simple and best

[@eric_bonucci_data]

Super clear, thanks a lot!

[@UnfoldDataScience]

Welcome Eric.

[@sarthakgarg184]

I have been searching for a better intuition on Gradient Boosting and this is the first video which gave me the best intuition. I am looking for research projects, can you help me with some topics on Machine Learning and Deep Learning which I could explore and ultimately go for a paper!
I'm also reaching out to you on LinkedIn for better reach. Thankyou for the video :)

[@UnfoldDataScience]

Thanks Sarthak, lets connect on LinkedIn and we can discuss more. Stay Safe. Tc.

[@mansibisht557]

Best video so far ! :') Thank you!!!

[@UnfoldDataScience]

Welcome Mansi.

[@MinaGholami-e2u]

Thank you. it was perfect explanation of gradient algorithm

[@UnfoldDataScience]

Glad it was helpful!

[@shikhar_anand]

Hi Aman, Thank you very much for the video. It was by far the clearest explanation for the topic. Just one doubt if you could clear it, How we can decide the number of iterations for any problem? You have iterated this for n=2, so how we can decide that.

[@UnfoldDataScience]

Hi Shikhar, We can pass it as parameter while calling the function.

[@krishnab6444]

thats a perfect explanation aman sir, in a simplest way, thanks alot sir, your videos are really helpful.

[@UnfoldDataScience]

So nice of you Krishna

[@vithaln7646]

this is very clear explanation ,

[@UnfoldDataScience]

Thanks Vithal. Happy Learning. Stay Safe!!

[@nurulamin7982]

Awesome.

[@UnfoldDataScience]

Thanks Nurul

[@manaspradhan2166]

Very well explained, Thank you

[@UnfoldDataScience]

Thanks Manas.

[@saravananbaburao3041]

One of the best video that I have ever watched for GB . Thanks a lot for the video. Can you please cover one video on Bayesian optimization . Really I find difficult to understand on that topic . Thanks in advance

[@UnfoldDataScience]

Noted.

[@sharmita220]

Thank you so much for all the videos. Its so clear

[@tradingdna5706]

Best teacher 👍🏻

[@tejaspatil3978]

Sir , it is very really best and very easiest explanation.
Wait for more videos

[@UnfoldDataScience]

Keep watching Tejas. Happy Learning.

[@Sagar_Tachtode_777]

Thank you for sharing such a piece of valuable knowledge in free.
May God bless you with exponential growth in the audience and genuine learners!!!

[@UnfoldDataScience]

So nice of you Sagar. Thanks for motivating me through comment.

[@anilboppanna]

Very nicely explained keep posting on such a quality videos..to unfold the data science Black box

[@UnfoldDataScience]

Thanks Anil. Happy Learning. Keep watching :)

[@soumyagupta9301]

I understood how Gradient Boosting works but still not understood why it works. Actually, I am not getting the intuition behind why we are interested in training the model on the residual error rather than the true value of y. Can you please explain this in a bit more detail? Anyway, I am a big fan of your teaching.,.it's so simple and easy to understand. Thank you for teaching so well.

[@UnfoldDataScience]

Thanks Soumya. you work with data more and you will know.

[@pratyushanand7429]

Great

[@ricardocaballero6357]

This is awesome, excellent explanation, thanks a lot

[@jude-harrisonobidinnu3876]

Very amazing videos. Concepts worth more than jumping into codes. Well done Sir!

[@adityasharma2667]

very well explained...i could say the best video to understand GB

[@UnfoldDataScience]

Glad it was helpful Aditya.

[@sachinmore8938]

You have got very good explanation skills!

[@MohitGupta-sz4bh]

How the algorithm decides the no of trees in Gradient boosting. And its advantages and disadvantages over Adaptive boosting. When to choose what... Please explain or reply in comments and yours videos are very helpful for someone like me who wants to switch his Career in Data Science field.
Also Can you please explain why we have the leaf nodes in the range of 8-32 in Gradient boosting and only one leaf node in Adaptive Boosting.

[@UnfoldDataScience]

# of trees - u can pass as parameter
AdaBoost vs GB which to choose - depends on scenario
I dont think there will be only one leaf node

[@shubhankarde4732]

great explanation...liked a lot

[@UnfoldDataScience]

Thanks a lot.

[@sandhya_exploresfoodandlife]

hi Aman.. your explanations are so good! thanks a lot

[@UnfoldDataScience]

Thanks Sandhya.

[@bhargavdr]

Twas very helpful thank you.

[@UnfoldDataScience]

Welcome Bhargav.

[@pokabhanuchandar9140]

Hi aman thanks for explaining the concepts. here I have one question for u "will ada boost accept repetitive records like random forest? "

[@shaelanderchauhan1963]

Please make numerical version of this by creating at least 3 iterations.

[@ravikiran1284]

Hi aman,
This is for gradient boosting and not for gradient boosting tree because in GBDT we update with gamma value at each leaf

[@preetibhatt5085]

Great explanation ... u said it right , couldn’t find right material for boosting on net . Could u pls make a video on XGBoost as well ??thanks for ur response in advance

[@UnfoldDataScience]

Sure Preeti.

[@ranajaydas8906]

The best explanation I ever saw!

[@UnfoldDataScience]

Glad you liked it. Continue watching

[@subhajit11234]

If you keep on growing the trees, it will overfit. How do you stop that? Will the model automatically stop ? or do we need to tune the hyperparameters? Also, it will be helpful if you can pick a record which we want to predict after training and demonstrate what will be the output, then that will be good. Going by your theory, all records you want to predict will have the same prediction. :)

[@UnfoldDataScience]

Hi Suvajit,
We must prune decision tree to avoid over fitting. Pruning can be done in multiple ways, like limiting number of leaf nodes, limiting branch size, limting depth of tree etc.
All these inputs can be passed to model when we call gradient boost. For optimal values, we should tune the hyper parameter.
Coming to part 2 of the question, all the records will not have the same prediction as error is getting optimized in every iteration. In the same model, If i try to predict for two different records, predictions will be different based on value of independent columns.

[@19967747]

Very well explained ! Please keep on making such nice videos !
Hope you reach 100k subscribers soon

[@UnfoldDataScience]

Thank you Ajinkya. Keep watching :)

[@firstkaransingh]

Very good and clear explanation 👍

[@alealejandroooooo]

This was great man, thanks!

[@UnfoldDataScience]

Thanks a lot.

[@sudheerrao9820]

Thank you for the video...very useful

[@UnfoldDataScience]

Welcome Sudheer, keep watching :)

[@pranjalgupta9427]

Awesome video 😍😍

[@UnfoldDataScience]

Thanks Pranjal. Hope you are doing great.

[@omarzahran4179]

Great Explanation,
but I want to ask two questions.
first Q:
why can't just make the update of the target value by:
The first iteration is ( base value + 1st Res pred )
The second iteration is ( (base value + 1st Res Pred) + 2nd Res pred )
The third iteration is ( (base value + 1st Res Pred + 2nd Res pred) + 3rd Res pred)
etc...
and if we keep doing that for like 10 iterations and take the output of the final iteration I think logically we should reach 100% accuracy!
why I used the learning rate! and why this model isn't the ultimate model with 0% error?
--------------------------------
Second Q:
why can't use this concept without predicting the residuals
ex: the first iteration is (base value + Res) and now I don't need any other model. it will end in just one iteration because simply the output will be my (prediction + the error) and of course will be = the target value
I'm pretty sure that this thinking is totally wrong because of some data leakage or something but I hope for an explanation of it.
and if this thinking is wrong why i can add the predictive residuals and can't add the residuals itself
thank you.

[@mdshihabuddin4099]

Thanks a ton for your spotless explanation . I have a question, how many residual models we will compute for getting our expected model, or how we will understand we need to compute that much of residual model.

[@UnfoldDataScience]

Good question Shihab, that number is a hyperparameter that can be tuned however there will be some default value for algorithm in R and Python.

[@mdshihabuddin4099]

Thanks for your response.

[@jagannadhareddykalagotla624]

@aman how to choose learning rate value and how to choose no.of trees

[@dimitarnentchev1107]

Very clearly explained ! Thank you.

[@UnfoldDataScience]

Glad it was helpful Dimitar!

[@kalluriramakrishna5732]

excellent aman. Thank you so much

[@peaceandlov]

Super Awesome mate.

[@UnfoldDataScience]

Thanks for the visit

[@aiuslocutius9758]

Thank you very much. Learning a lot from your videos!

[@UnfoldDataScience]

Welcome.

[@hirdeshkumar4069]

How will you define you learning rate and how did you arrive to the value 0.1?

[@UnfoldDataScience]

This is just a number I took for explanation however this is a parameter that can be tuned..

[@mamatha1850]

clearly explained.thanks bro

[@UnfoldDataScience]

Happy to help Mamatha :)

[@sameerpandey5561]

Thank you Aman..It was very crisp and clear...explanation
Just a request..Please add a Video explaining GBDT in case of classification problems ...That would be very much helpful :-)

[@UnfoldDataScience]

I will try my best.

[@ajaykushwaha-je6mw]

Thank you so much for such a nice video. It help me to understand thee concept of GB Algo.

[@UnfoldDataScience]

Glad it was helpful Ajay. Happy Learning!

[@prasadjayanti]

good work..

[@UnfoldDataScience]

Thanks a lot.

[@bhushanchaudhari378]

Very well explained sir🎂.. thanks a ton

[@UnfoldDataScience]

Welcome Bhusan. Keep watching :)

[@anirbansarkar6306]

It is always great to learn from your videos. I have one small doubt:
Stem acts as the basic unit in adaboost. But if we change the algorithm from decision tree to say logistic regression, then also do adaboost uses stem as basic unit (as it is tree) or something else.

[@UnfoldDataScience]

Hi Anirban. i dont think u can use any other weak learner apart from DT in sklearn GB.

[@SuperShiva619]

Ur awesome🙂 the intent at which u explain concepts are mind blowing.

[@UnfoldDataScience]

Thaks Satish, keep watching :)

[@sankararaoyenumala8737]

tq sir,its good explination.

[@devendrakumarks4859]

This was very well explained brother. Can you also please bring in the classification problem explanation? I mean how the initial base prediction is done and the other things? That will be of a good help.

[@UnfoldDataScience]

Thanks Devendra . Will do. Happy Learning. Tc

[@kemarwhittaker5683]

Awesome video

[@UnfoldDataScience]

Thanks a lot. Stay Safe. Tc

[@chaitanyakaushik6772]

Excellent explaination sir.

[@UnfoldDataScience]

Thanks and welcome

[@taoris603]

Do u have explanation for Gradient Boosting Classifier? Or pls make it, thank u so much

[@samruddhideshmukh5928]

How does the gradient boosting stops or when does it stop?(Does it stop when the loss becomes minimum or do we specify n_estimators for it to stop?)
Also pls explain gradient boosting for classification if possible..it would b every helpful

[@UnfoldDataScience]

Stopping is based on model hyperparameter

[@goelnikhils]

Amazing Content. Thanks a lot

[@UnfoldDataScience]

Welcome Nikhil,pls share with friends

[@pritampawar6478]

very well explained