The intuition for denoising auto encoder is greatly presented by adding noise to original image and recreating the image with ae 💯

@23:59 Number of trainable params in decoder are 2. One for w1cos(z), one for w2sin(z) respectively.

53:34 trying to answer what happened here, after 4 rounds of rotations (linear transformations) and squishing (with ReLu) the final affine transformation lead to the projection of points on a straight line. Which still is not linearly separable space, missing my intuition on what's special about this piece🙃

Alfredo thank you very much for this concrete explanation. Yann's lectures are full of dense knowledge and I had some doubts and gaps that are filled by this practicum. Appreciated/

Guessing, the five lines at the end of video are basis vectors in un warped space or transformed space where we can draw lines to separate the colored dots. Support vector machines applies the same principle but with pre-defined functions and they are not as sensitive as neural nets.

Thanks alot for this.

1:57:06 What Yann LeCun did in 1989 without Python, Pytorch or Jupyter Notebook is clearly spectacular.

Starting this playlist on 3rd October 2024, Extremely thankful for open source education. I'm excited as well as afraid of the mathematics involved.

Hi Alfredo, thank you for your content! Is there a place where we can find the first part of this course? I would like to take the whole course while I go through the book as well 😊

Lewis Carol Thompson Brenda Allen Laura

Mass 🎉 video

[NYUDLFL24-alumni] 1:02:47 Question about the dimensionality of weight matrix W_g; We say that the dimension of W_g should have dimension [d_f x d_g], since the transpose of W_g has dimension [d_g x d_f]. But since z_g is dimension [d_g x 1], z_f is dimension [d_f x 1], isn't the forward pass given as: z_g = W_g * z_f? If so, shouldn't W_g have dimension [d_g x d_f]? --> [d_g x 1] = [d_g x d_f] * [d_f x 1] Instead of [d_f x d_g]? Unless the forward pass should be z_f^T * W_g ?

Yann is awesome.

Great resource, Text needs to have a background to be eligible

Gelatos Alfredo! I use to eat glasses at Fredo in Argentina. Atcold. Thank you so much Alfred for sharing all this stuff. You are making history. Gracias!

9/11/2024 Update: Answer given by ChatGPT 4O to the example question given by Yann at 1:35:58 - "what is the country that has a common border with Germany with the largest commercial exchanges with China?": Poland.

Thanks, I am also learning something new!

Thanks!

The part where he mentioned FMA made me burst out laughing! Can't wait to join in-person classes!

I can totally see how a quantum computer could be used to perform gradient descent in all directions simultaneously, helping to find the true global minimum across all valleys in one go! 😲 It's mind-blowing to think about the potential for quantum computing to revolutionize optimization problems like this!

Thank you so much, it's amazing how easy you make this to understand.

Thank you, Alfredo! I just started this course, and I think it's incredibly detailed and amazing. I can't express my excitement in words! ❤️❤️❤️

Amazing ❤❤❤❤❤

Just finished all videos. Really amazing. Thank you for sharing.

J'adore la couverture de Fluide Glacial avec Super-Dupont ... on est vraiment de la même génération !

Hey I’m kinda new to this field , but I’ve tried to catch up Which of the playlist do you recommend me to follow ? The one with Dr Yann Lecunn or the current one ? 🙏

This ASMR is a bit odd.

Why cant we use counters for the loops in neural nets - would a loop not make the network more robust in the sense of stabilizing output?

only a few words on his own masterpiece work haha

Hello! First of all, thank you for uploading the material. Very, very good course. However, in this part of EBMs, I'm a little bit confused: Lets supposed that I've trained a denoising AE (or other variation) with a bunch of y's. After training, how do I use it in practice? I'd pick a random z and use to generate a y_tilde? From which distribution I'd sample such z?

Thank you, Alfredo! :)

❤

Thank you, Alfredo. You made such a great visualization!

How were the neural nets trained before 1985 ie before back prop was invented?

Thank you, Alfredo :)

Amazing Lecture!

Thank you Alfredo, you have made a very clear explanation of this topic. :)

Thank you, Alfredo!

This principle of running differential equations backward is used in diffusion when you find the lagrange Loss function from the Score which is the time reversing Langevin dynamic equation. Cost and Energy or Momentum and Energy. Both are deterministic reversible dynamic systems.

Thank you, Alfredo! I am happy that you are back

Hello, Alfredo! :)

Thank you for sharing. I'v one question about the NNs on scrambled data. If I had to make a prediction I would have said we will have an accuracy about 15%, not more, thanks to the amount of pixel that can help to determine which digit is corresponding. So is it enough to get an accuracy of 83-85% or there is something else? I supposed that the fully connected neural network would have duplicate the filters, but there is no change with the scrambled data.

Thanks a lot! Your content is giving me motivation to get back into this field. Keep it up please 👏🙏

Thank you for posting this. Looks like a great hour is ahead.

@14:04 Paraphrase: Missing a positive (i.e., false negative) is more critical (i.e., worse) than a FALSE POSITIVE. (Note: "falsely identify a negative case" means "falsely identify AS A POSITIVE what is actually a negative case.)

Loved the video ❤. hey i am working as an nlp engineer in india. do you have any remote opportunities for me? let me know if you have something.

that's the entire reason to teach :) learn how to do something and pass it on

@5:24 Paraphrase: "So, what is the accuracy of a classifier that classifies everything as HAM, detecting no SPAM, thus yielding NO POSITIVES ?"

Thank you so much for sharing this series. I especially loved the vintage ConvNets and the brain part :) I have a question: I didn't understand how we define the number of feature maps. For example, in 1:27:00 , how did we go from 6 feature maps in layer 2 to 12 feature maps in layer 3? (By the way, there are 16 feature maps in layer 3 (C3) in the architecture of LeNet-5 in this paper: yann.lecun.com/exdb/publis/pdf/lecun-98.pdf (Fig 2. the architecture of LeNet-5).

*Summary* *Probability Recap:* * *[**0:00**]* *Degree of Belief:* Probability represents a degree of belief in a statement, not just true or false. * *[**0:00**]* *Propositions:* Lowercase letters (e.g., cavity) represent propositions (statements). Uppercase letters (e.g., Cavity) are random variables. * *[**5:15**]* *Full Joint Probability Distribution:* Represented as a table, it shows probabilities for all possible combinations of random variables. * *[**10:08**]* *Marginalization:* Calculating the probability of a subset of variables by summing over all possible values of the remaining variables. * *[**17:04**]* *Conditional Probability:* The probability of an event happening given that another event has already occurred. Calculated as the ratio of joint probability to the probability of the conditioning event. * *[**16:14**]* *Prior Probability:* The initial belief about an event before observing any evidence. * *[**16:40**]* *Posterior Probability:* Updated belief about an event after considering new evidence. *Naive Bayes Classification:* * *[**32:48**]* *Assumption:* Assumes features (effects) are conditionally independent given the class label (cause). This simplifies probability calculations. * *[**32:48**]* *Goal:* Predict the most likely class label given a set of observed features (evidence). * *[**44:04**]* *Steps:* * Calculate the joint probability of each class label and the observed features using the naive Bayes assumption. * Calculate the probability of the evidence (observed features) by summing the joint probabilities over all classes. * Calculate the posterior probability of each class label by dividing its joint probability by the probability of the evidence. * Choose the class label with the highest posterior probability as the prediction. * *[**36:24**]* *Applications:* * *Digit Recognition:* Classify handwritten digits based on pixel values as features. * *[**47:34**]* *Spam Filtering:* Classify emails as spam or ham based on the presence of specific words. * *[**33:56**]* *Limitations:* * *Naive Assumption:* The assumption of feature independence is often unrealistic in real-world data. * *[**42:11**]* *Data Sparsity:* Can struggle with unseen feature combinations if the training data is limited. *Next Steps:* * *[**1:05:58**]* *Parameter Estimation:* Learn the probabilities (parameters) of the model from training data. * *[**59:53**]* *Handling Underflow:* Use techniques like logarithms and softmax to prevent numerical underflow when multiplying small probabilities. i used gemini 1.5 pro to summarize the transcript