Відео

Very good lecture, Andrej. Also, the student who interrupted Andrej by saying 'that was what I said' didn't have a good tone and annoyed me.

hey internet, i was here

andrej has the most unique way of telling things and generally they are more intuitive, what a maniac!

Just want to say a big Thank you, this course had helped me prepare the foundation for my PhD back in 2017. Literally watched it multiple times to internalise the concepts

and this is the 100th

this is the 99th comment

27:40

6:00

Amazing Course ! Thank you!

Started my journey and watching this course now. Couldn't attend Stanford University but still good to be in the class in 2024 : )

This is helping me quite a lot, thanks!!!

In 2015, the world didn't know about the problem of batch normalization

1:14:27 - The network never fully converges, but at some point you stopped caring. Because it has been 2 weeks and you are just tired. 😅

You are one of the best Andrej. You make learning so fun, with moments like 27:45 😄 Forever grateful.

This clears lot of doubts I had in my head. Thank you Andrej.

Can't thank Andrej, the cs231n team, Stanford enough. Thoroughly enjoy your lectures. Knowledge is one form of addiction and pleasure and thank you so much for providing it freely. I hope you all enjoy giving it as much as we enjoy receiving it.

1:09:10 it worked :)

does he speaks 1.5x by default 😛

At 39:00, what did he mean by jiggling the scores?

one of the best videos on RNN who explains the code perfectly from sratch.

This content here, is GOLD.

28:00

I love you Andrej ❤

This is a beautiful lecture - gave a very fundamental understanding of backward propagation and its concepts - I see backward propagation correlates to demultiplexing and forward prop corresponds to multiplexing where we are multiplexing the input .

haha. at 1.17.00 Justin says "it's cool but not sure why you would want to generate images"... 🙂

Seems to be a mistake in the "Computing Convolutions: Recap" slide - it says "FFT: Big speedups for small kernels" when it should be "big kernels"?

YOLO is king now hahaha.

when this video was recorded I could not even factor a quadrtic equation. Now I can watch this and follow the math w realetive ease. wow

Day -2 . 3 lectures and counting...

10

9

8

It's still a good way to begin the journey of CV in 2023 , thanks for teaching

Es muy interesante

Tarea 8

Can someone explain why the variance is reduced by 1/2 when using ReLU? Take for example a sample of size=100_000 from a normal distribution N(0, a), pass it through the ReLU and then calculate its variance. Would it be a/2? Moreover, on 45:30 why by setting larger weights the distribution of activations changes shape compared to when using Xavier? I am expecting a flatter distribution compared to Xavier, but not that shape with these peaks on the boundaries. Finally, how these distributions of activations are calculated? Passing many samples through the network with fixed weights?

I wanted to code all that up like you did but I am not able to do it 🥲

i am come from his blog, and he say he put a lot of effort for this class.

Thanks internet

Question! (from an absolute beginner): For the slide shown @24:52, (The bottom example on "Backward pass: guided bakpropagation") Why don't all the negative numbers map to 0? My understanding was that it should automatically be done to deconstruct the image properly. Thnx.

Awesome as always! Is notes available to general public? If so, where do I find it?

She said that she is giving @Andrej a challenge and ChatGPT4 did the same thing in the 14 March 2023 release .

This is the course, I finally understood it. Thank you so so much!!!!

37:00 wrong, each of the four images are a different set of hyper parameters for regularization. As for different initialization - you should look at the paper at figure 4 - The 9 images are different initializations.

37:49 they ask professor why do we look for the minimum of the -log(p) instead of just looking directly for the maximum of p. He says that it makes the math more convenient but I thought the reason was that you want to maximize the joint probability p1*p2*p3*…*pN for your N samples. So instead of maximing a multiplication, you maximize the sum of the logs: log(p1)+log(p2)+log(p3)+… +log(pN).

about reconstruction: reconstruction vs running throw the NNW <=> integral solving vs derive in the sense that every node that was activated gives more information then any node that was not. x^2 + x +15 dx => 2x + 1 => x^2 + x + n where n is the missing info resulted from no surviving the the boundary function.

ty very much!!!

These lectures are amazing. I still go back to review these every few years. Thank you sir.

1:10:10 “can we give a computer one picture and outcomes a description like this… you give the computer one picture - it gives you one sentence… we are not here yet”. This is 2016, and Dr. Fei-Fei Li cannot envision even in her wildest dreams that a few years later, we shall not only overcome this problem but solve the opposite - give a computer a sentence, and it shall create a picture. She was too ‘shy’ to even ask a computer to do such a thing. What is this something that we don’t even dare to ask a computer to do today, but in 2028 computers will do? 😊

Did you hear about the mathematician mountain climber? He got stuck at a local maximum