Hii i can help u for reaching monitization for ur channel

Thanks, that was a great video!

Good explanation! Thank you so much!

There is an error in the 1st slide. Entropy of the fair coin equals 1 bit, not 0.7. It's probably natural log was used in this graph.

Really enjoyed the guitar tuning analogy. Can’t stand playing out of tune guitars or when the intonation is slightly off.

Is Stock price prediction an suitable real world use case for this Online learning ?

slide 13 is a little bit confusing

the last example was a little bit confusing.

Appreciate it 👍

Hi，professor ! I have a question: I remember in previous video I learned that random forest can deal with missing data by surrogate splitting, but now in this video professor said proximities can be used for imputation. I'm confused. If random forest can deal with missing data then why do we need imputation? And in this video in imputation part, step one is using median to impute the data, why we need to impute if random forest already can use surrogate splitting to get the prediction result? why dont we use the result generated by surrogate splitting to compute proximity to impute the data and keep doing step 2 and 3? Besides, I feel I dont totally understand how does surrogate splitting work. I searched on the internet, but I still dont know the details. I just know it is about finding another variable that can generate the good result as variable[the primary split] with missing values. But how does the calculaton happen to make us know that the variable with missing values can be the best split at specific point[and if we already know that the variable with missing values can be a good split at specific point then why do we need surrogate splitting]?

Absolute gold

great video thanks, very clear

thank you

how do we get phi values in step 4?

LET'S GOOOOOOOO 💫💫 THANK YOU FOR TAKING YOUR TIME TO MAKE THESE VIDEOS💯💯💯💯❤❤

Is this algorithm inspired by k-means clustering

Thank you! You explained thing very clearly.

Very helpful video. The visualization in the OOB is very easy to understand. Thank you!

great lecture! what if some of the variables to be optimized are limited in a certain range? using multivariate normaldistribution to generate offspring might exceed the range limit?

Excuse me, but wouldn't z1+z2+z3+....zT be (-1)^T/2 instead of (-1/2)^T? Anyway, you did a great job. Congratulations!!

first :)

It's so hard to hear what you're saying, please amplify the audio post processing on your future uploads. Excellent presentation nonetheless, you explained it so simply and clearly. <3

at 4:00 isn't the square supposed to be inside the square brackets?

HI, great course, however a small note, at 12:20 I think that the function on the left might not be convex, as the tangent plane in the "light blue area" is on top on the function, and not below the function, thus violates the definition of convexity (afaik they are called Quasiconvex function that sort of functions)...

finally found out clear logic behind the weights. Thank you so much🎉

Nice video. I have a question. How to calculate shapley value for classification problem?

Thank you for doing this video in Numerator layout. It seems many videos on machine learning use Denominator layout but I definitely prefer Numerator layout! Is it possible you could do a follow-up video where you talk about partial derivative of scalar function with respect to MATRIX ? Most documents I've looked at seem to use Denominator layout for this type of derivative (some even use Numerator layout with respect to VECTOR, and then switch to Denominator layout with respect to MATRIX). I assume it's because Denominator layout preserves the dimension of the matrix, making it more convenient for gradient descent etc. What would you recommend I should do?

💐 Promo'SM

So, the permutation order was only to define which feature will get the random value? Not creating a whole new instance with feature order same as the permutation order? (The algorithm shows S, j, S- , but your example shows S, S-, j)

Thank you!

Nice lecture. I also recommend Dr. Ahmad Bazzis convex optimization series.

Great work

Hi, sadly your github link doesnt work for me. Thanks for the video.

Can you explain more about the ommitted variable bias in M-plots? My teacher told me that you can use a linear transformation to explain the green graph by transofmring x1 and x2 to two independent random variables x1 and U. Is that true?

Thank you so much for this video. Consider example on 7:20 min. -- but doesn't it look like feature permutation? Shouldn't I use expected values for other variables (x2,x3) ? Thanks in advance.

Thanks

the material is not accessible right now, can someone reupload it ?

what do you mean with "pessimistic bias" ?

good explanation madam

Good Explanation Madam

Hallo, I'm a student in TU dortmund, our lecture also use your resources, but the link in the description is not working now, how can we get access to the resources?

Great Explanation!! Thank You

Excellent. Thanks

Finally some serious content

Thanks Sir, but excuse myself (zero statistics & mathematics background), but what does this video suggest about using Brier Score for measuring forecast?

Nice explanation! You are using the same set of HP configurations λi, with i=1, ...,N through fourfold CV (in the innner loop). But, what happend if I would like to use a Bayesian hyperparameter to sample the values of the paramters? For example, for each outer cv with its corresponding inner cv, could I use a Bayesian hyperparameter search? Then the set of HP configuration wouldn't be the same in each inner cv, and then the question is: Can be the set of HP configurations different in each inner cv and is it still valid this nested cross valitadion method?

awesome!! thanks so much!

thank you, well explained.

Heres a free resource on one of the most important academic concepts of the modern age: 800 views. GG humanity. GG

thank you for the explanation