@OliverJanShD
Thank you for posting these! I find your content to be the most educational/entertaining on this subject matter.
@sans8119
Would be amazing if we could have similar quality content on Reinforcement learning from you! Thank you for having the lectures on Youtube!
@rohit2761
Kilian + StatQuest = 100 percent learning. Thank You Prof Kilian, absolutely loved the series. Now I'll head over to Deep Learning :)
@linxingyao9311
Really love it, I learn quite a lot each time I watch this video.
@HLi-pc4km
"Boosting is gradient descent in function space" - I am gonna steal this line.
@rishabhkumar-qs3jb
Awesome lectures, you made machine learning easy for me, now I have better grip over the concepts of boosting .
@vatsan16
"It begins with a Tay and ends with a Lor" hahaha "Taylor expansion" "That's right! HOW DID YOU KNOW?" i burst out laughing at that.
@vatsan16
I have had my fruit and I am ready!! Lets do this!
@andreamercuri9984
the idea of putting the labels in a very high dimensional space is pedagogically very, very good! I had some troubles trying to understand AnyBoost from the original paper, but Prof. Weinberger made the principle behind it very easy to understand. Great as usual
@ruman2494
Explained very well. Thanks a lot sir.!
@juliocardenas4485
It’s wonderful to revisit these lectures
@xiaoyu5181
Your lecture really helped me understand bagging and boosting more clearly. Thanks for your sharing.
@kirtanpatel797
Waking up from coma, What is boosting? Drunk Guy.
@Psingh8077
Great Lecture Series
@KW-fb4kv
Your English is so perfect.
@HLi-pc4km
I have to leave another comment. Prof. Weinberger is a teaching genius. Also, who says germans ain't funny?
@Ange-ClementAkazan-k5s
Very Great Lecture
@yannickpezeu3419
so well and funnyly explained TY
@aciobanusebi
Really great lectures! Thanks! I have a question: In Gradient Boosting the algorithm says "h* = argmin sum (h(x_i) - t_i)^2". I don't understand if one should: 1. go through all the possible regression trees (with a maximum depth set) and find that "h*" OR 2. one could just apply CART with the new labels "t_i" instead of "y_i". As I might guess, we should go with 2. But doesn't 2 give just an approximate solution to that maximization problem? Am missing something?
@ayushinayak3980