The formula for the neural network in this video should be a 2 level summation instead of one level.
@KittyCat-lp3zyАй бұрын
یاشا آذربایجان ثروتی ❤
@MonkkSoori3 ай бұрын
At 20:20 why does Phi(Q_i) not cancel out in the numerator and denominator?
@janesun90083 ай бұрын
Thank you for sharing this lecture, prof. Great quality and easy to understand!
@NavaAbdolalipour3 ай бұрын
من با شما قلمچی اردبیل بودم،بعد سالها توی کشورهای نزدیک بهم اسمتون رو دیدم،خوشحالم از موفقیت های هم دوره ایی ها
@simaranjbari3 ай бұрын
your explanation was very nice and easy to understand. Thank you!
@fierydino94024 ай бұрын
Wonderful lecture!! Thank you for sharing
@PradeepKumar-zy6cd4 ай бұрын
Can you please share the slide
@prabhavkaula96974 ай бұрын
Thank you for the lecture! ☺️
@ax53444 ай бұрын
@1:58:57 You said you will explain different procedures to generate different responses later. I did not find it till you start discussing Step 3. Could you illustrate further?
@ax53444 ай бұрын
@2:15:30 Found it. Thanks!
@ax53444 ай бұрын
When you upload the video, could you set the speed to 1.5? Right now, I'm setting it to 2X, it is still very very slow.
@sabujchattopadhyay4 ай бұрын
Can you please share the slides? (2)
@miquelnogueralonso25764 ай бұрын
Can you please share the slides
@bardiasafaei4574 ай бұрын
Thank you Soheil for the great content and the clear way of explanation! Could you also share the final written notes of each session for download?
@ai__765 ай бұрын
Massage lesson! Tnx
@amiltonwong5 ай бұрын
Thanks a lot for providing such an excellent lecture. Would it be possible to release the notes for study? Thanks~
@parhamsalar38265 ай бұрын
Many thanks for your excellent lectures, particularly those on diffusion models. I do have a few inquiries regarding models of conditional diffusion. Can we think of text vectors as the query (Q) and image vectors as the key (K) and value (V) in cross-attention instead of image vectors as the query (Q)?
@parisaemkani57305 ай бұрын
Hi, could u please introduce a good course in the basics of machine learning and deep learning for beginners?
@Stealph_Delta_30035 ай бұрын
Thanks for sharing.
@sdiabr67925 ай бұрын
Real quality content
@mozhganmomtaz81695 ай бұрын
I just want to thank you 🤗
@INSTIG8R5 ай бұрын
This here is the best video on SWIN transformers
@MrNoipe5 ай бұрын
The handwriting is difficult to read, maybe write slower or with a different brush?
@naeemkhoshnevis5 ай бұрын
Thanks for uploading these lectures.
@miladkhademinori27095 ай бұрын
❤
@naeemkhoshnevis5 ай бұрын
Thanks for uploading the lectures.
@Nerraruzi6 ай бұрын
Thanks so much for sharing this updated version of the course!!
@shayanmohammadizadeh1726 ай бұрын
It's minute 30 of the video and I have watched +8 ads. Really attention is all we need!
@Umar-Ateeq4 ай бұрын
you can use "adblock for youtube" extension to avoid ads.
@junqi70506 ай бұрын
Thank Soheil for sharing the updated deep learning theory courses. I ever followed Sohail's former lectures in 2020, where I learned the theoretical knowledge of deep learning in terms of representation, generalization, and optimization. I found that Soheil's course schedule this year has substantially changed to state-of-the-art transformer-based technologies, such as large language models, etc. I plan to catch up with Sohail's updated deep learning foundation course this year and really appreciate the new lecture videos.
@mohammadshahbazhussain20296 ай бұрын
Thank you for sharing it
@user-pz5nn2kg2j6 ай бұрын
Hi professor, I was also wondering that if you plan to to add some contents related to distanglement learning? like nonlinear ICA which I think is very theoretically interesting and important.
@user-pz5nn2kg2j6 ай бұрын
Sorry, there's a typo. It should be 'disentanglement'.
@user-pz5nn2kg2j6 ай бұрын
Thanks for updating this really amazing course. I've read the syllabus of this semester, and find it is really interesting, especially the generative models and multi-modal models part. Hope to see more latest course videos. Thanks a lot for your effort of sharing the contents of this amazing course.
@hesamce6 ай бұрын
Thank you for sharing the updated version of the course🙏
@AyushSharma-ie7tj Жыл бұрын
Really nice lecture with a very even pace. Thank you for sharing.
@StratosFair Жыл бұрын
Great lecture. Thank you for sharing
@hamedgholami261 Жыл бұрын
explanation of: "Loss landscapes and optimization in over-parameterized non-linear systems and neural networks"
@mojtabakolahdoozi2418 Жыл бұрын
Great lecture on the highly ignored ground! thanks
@quanguyenang1615 Жыл бұрын
Thanks for the great lectures, Prof. Soheil.
@sylus121 Жыл бұрын
25:00 (Bookmark)
@Thaumast Жыл бұрын
24:18 The loss function is sometime defined by an L and sometime edfines by the caligraohic L, are they the same? thank you very much !
@bryanbocao4906 Жыл бұрын
42:30 one option could be KL divergence loss?
@mskang009 Жыл бұрын
Such a great lecture I've seen in KZread related to self-supervised learning. So many thanks!
@sinaasadiyan Жыл бұрын
great explanation, just Subscribed!
@sumitsah6092 Жыл бұрын
How can we guranttee that w_t lies within the ball??? Because if that is not the case then we can't apply PL inequality. Please comment.
@zonghua Жыл бұрын
The handwriting is unclear.
@user-cp8uy9om7o Жыл бұрын
Amazing video. Thanks!
@luisjacobson4375 Жыл бұрын
✌️ ρ尺oΜ𝐎ᔕᗰ
@jfjfcjcjchcjcjcj9947 Жыл бұрын
Thanks for the lecture! I think that there are a couple of points that would make the lecture more digestible if they are explained a bit more in depth. First, is the definition of interpolation, different people imply different things when referring to interpolation. I presume that Misha here by interpolation refers to the ability of estimators to achieve zero training error and still not overfit to the hold-out test set? The second aspect which might require some more explanation is @40:05 when Misha is describing the difference between interpolation accuracy and interpolation in the l2-norm. I presume that what is meant by l2-norm interpolation is that 2 estimators with the same l2-norm on their parameters exhibit the same generalisation capabilities, is that right?
Пікірлер
What is true distribution?
Thank you, Sir.
The formula for the neural network in this video should be a 2 level summation instead of one level.
یاشا آذربایجان ثروتی ❤
At 20:20 why does Phi(Q_i) not cancel out in the numerator and denominator?
Thank you for sharing this lecture, prof. Great quality and easy to understand!
من با شما قلمچی اردبیل بودم،بعد سالها توی کشورهای نزدیک بهم اسمتون رو دیدم،خوشحالم از موفقیت های هم دوره ایی ها
your explanation was very nice and easy to understand. Thank you!
Wonderful lecture!! Thank you for sharing
Can you please share the slide
Thank you for the lecture! ☺️
@1:58:57 You said you will explain different procedures to generate different responses later. I did not find it till you start discussing Step 3. Could you illustrate further?
@2:15:30 Found it. Thanks!
When you upload the video, could you set the speed to 1.5? Right now, I'm setting it to 2X, it is still very very slow.
Can you please share the slides? (2)
Can you please share the slides
Thank you Soheil for the great content and the clear way of explanation! Could you also share the final written notes of each session for download?
Massage lesson! Tnx
Thanks a lot for providing such an excellent lecture. Would it be possible to release the notes for study? Thanks~
Many thanks for your excellent lectures, particularly those on diffusion models. I do have a few inquiries regarding models of conditional diffusion. Can we think of text vectors as the query (Q) and image vectors as the key (K) and value (V) in cross-attention instead of image vectors as the query (Q)?
Hi, could u please introduce a good course in the basics of machine learning and deep learning for beginners?
Thanks for sharing.
Real quality content
I just want to thank you 🤗
This here is the best video on SWIN transformers
The handwriting is difficult to read, maybe write slower or with a different brush?
Thanks for uploading these lectures.
❤
Thanks for uploading the lectures.
Thanks so much for sharing this updated version of the course!!
It's minute 30 of the video and I have watched +8 ads. Really attention is all we need!
you can use "adblock for youtube" extension to avoid ads.
Thank Soheil for sharing the updated deep learning theory courses. I ever followed Sohail's former lectures in 2020, where I learned the theoretical knowledge of deep learning in terms of representation, generalization, and optimization. I found that Soheil's course schedule this year has substantially changed to state-of-the-art transformer-based technologies, such as large language models, etc. I plan to catch up with Sohail's updated deep learning foundation course this year and really appreciate the new lecture videos.
Thank you for sharing it
Hi professor, I was also wondering that if you plan to to add some contents related to distanglement learning? like nonlinear ICA which I think is very theoretically interesting and important.
Sorry, there's a typo. It should be 'disentanglement'.
Thanks for updating this really amazing course. I've read the syllabus of this semester, and find it is really interesting, especially the generative models and multi-modal models part. Hope to see more latest course videos. Thanks a lot for your effort of sharing the contents of this amazing course.
Thank you for sharing the updated version of the course🙏
Really nice lecture with a very even pace. Thank you for sharing.
Great lecture. Thank you for sharing
explanation of: "Loss landscapes and optimization in over-parameterized non-linear systems and neural networks"
Great lecture on the highly ignored ground! thanks
Thanks for the great lectures, Prof. Soheil.
25:00 (Bookmark)
24:18 The loss function is sometime defined by an L and sometime edfines by the caligraohic L, are they the same? thank you very much !
42:30 one option could be KL divergence loss?
Such a great lecture I've seen in KZread related to self-supervised learning. So many thanks!
great explanation, just Subscribed!
How can we guranttee that w_t lies within the ball??? Because if that is not the case then we can't apply PL inequality. Please comment.
The handwriting is unclear.
Amazing video. Thanks!
✌️ ρ尺oΜ𝐎ᔕᗰ
Thanks for the lecture! I think that there are a couple of points that would make the lecture more digestible if they are explained a bit more in depth. First, is the definition of interpolation, different people imply different things when referring to interpolation. I presume that Misha here by interpolation refers to the ability of estimators to achieve zero training error and still not overfit to the hold-out test set? The second aspect which might require some more explanation is @40:05 when Misha is describing the difference between interpolation accuracy and interpolation in the l2-norm. I presume that what is meant by l2-norm interpolation is that 2 estimators with the same l2-norm on their parameters exhibit the same generalisation capabilities, is that right?