Video

35:30 should be dx instead of dt

Doubts: 1. If i generate random number plot them I get the pdf and when I cummulatively plot it I get the cdf. Now you are saying that this new cdf will always be uniform. But from theory we know that the cdf of exponential distribution is not uniform . I am really confused now

Can anyone please explain how do I find the distribution of f(x) in the very first place?

My major question is: Given just some data points about an experiment or prior knowledge how do we find that what kind of distribution to use?

In the marble flipping example how did we plot the distribution in the very first place. Did we do number of trials for it? Sowmthing like: Trail 1: picked 10 marble : got 6 blue 4 not blue Trial 2: picked 10 got 2 blue 8 not blue Found the poportion of this And then the number of trails for each proportion plotted as density? Is this correct understanding?

Why is the likelikhood not 10C5 p^5 (1-p)^5 : (p ~ theta) the batsman can hit the ball in any sequence. Why don’t we consider this for calculation?

Great course, Prof Chen. My only advice is please create playlist for each course. Otherwise you have to search for each video separately :(

Your classes are great, as always. However, I believe your audience would be much larger if you used more informative titles on your CZcams videos. You know, just use some keywords like "R", "tidymodels", "how to create predictive models using R", etc

Bro i love this prof and i havent even taken his class yet

Beautifully explained. Thank you so much

useless :(

Hi.. Nice work. Could you share this R code?

Good video. It would have been better without the distractions. The animation was great. I have used N-M to optimize 15 parameters using REAL data. Often methods that use derivative, Jacobians and Hessians don't work with real data because the data is noisy. What I wonder about is why isn't N-M used instead of all the gradient methods I see that don't work well with real data. I have spent a lot of time analyzing my data and it is not like the simple 'bowl like" landscapes used in YT videos. It is more like going down the Colorado river in the Grand Canyon where steps of any size will run into a "canyon wall" in some dimension. The path to minimizing is very narrow. N-M seams to skip all that. The only reason why N-M isn't efficient is because it collapses so the simplex is small. At this time another method could be tried to finish off the last part of the optimizations because the steps would be small.

Very good explanation. Thank you!

Why dont calculate all of these points and spot the minimum leading strategy?

much clearer than my lecturer lol, thank you so much

Real good explanation and the animation at the end made it even more amazing. Those having problem with background voices should try switching off the stable volume option it kind of equalises the background noises😅

AMAZING!!!

Thanks you so much for your help!

what a boring lecture !

Great video! I guessed the wrong view quiz answer :/

Great video! I guessed the wrong view quiz answer :(

Great video! I guessed

Great video! I guessed the right view quiz answer :)

Great video! I guessed the wrong view quiz answer :(

Thanks Miles!

This was brilliant thank you. Both examples were nice and intuitive. I am trying to understand this stuff because I will be using FiM for optimizing dose selection for a clinical trial at some point in the next year. I liked the linear model because it highlighted that if your linear model exhibits a large degree of noise, then a larger degree of variability in your input variable is needed to appropriately estimate the output variable with sufficient precision. I'll leave this little example here below. For Sigma = 2, and X = [1,500,1000]. I(THETA) = X(transpose)*X/Sigma^2 = 1,250,001/4 = 312500. However, if you only had X values 1,2 and 3, I(THETA) drops down to 14/4 = 3.5. The same intuition applies to non-linear models where the ability to fit them is going to be heavily dependent on the spread of the data.

What a genius of a method

Thanks for the explanation!

Child talking, phone buzzing, animals yipping. It’s like you went for the how not to make a video achievement.

Great lecture!

This is great, thanks for posting.

How did you do that

very helpfu thanks!

Great video Dr. Chen!

Great lectures! Thanks for sharing!

🐐

This really helped me understand the Inverse Transform Method more than any proof I've seen. It just really helps when it's explained conceptually before proving mathematically. Keep up the good work! <3

This is the most underrated channel ever!!! wow! great stuff! Thank you Miles!

Thank you Thank you!🙏👌

Thank you so much!🙏👌

Great explanation! Thanks a lot

Very good explanation! Thank you!

I have problem, when i install tinytex, the command stop at Running fc-cache -v -r > It didn't show any installation progress nor message of installation successfull (i wait 15 minute ,my machine is not slow pc)

Great video! Thank you very much!

Could we adapt NM to use sphere instead of simplexes? My problem is that simplex exists only in Euclidean Geometry.

This is great. Thank you so much!! Do you have any video/lecture where I can see how to run this for spatial data? Many thanks in advance!

Thank you for this amazing resource.

nice vid prof!!

You have an amazing account!! I wish I had found it earlier😢😢❤️❤️