Hi, I am a student from a really poor country who really loves to learn about deep learning, and thanks to you it is now accessible to me. I am really grateful. thank you.
I hope that you understand the great value you're giving! I'm truly pleased watching your videos. You are giving us exposure for a lot of cutting edge technology/information/Concepts that is not available at our universities (or at-least when i was at University). Thank you for the hard work! If I can send you a picture it would be me with a paper and a pen, taking notes and studying your material.
Aww, that's adorable 😊 Yeah, I think I identified a gap in the current educational resources. Hopefully, I'll do justice to the task I've embarked and to the role I've taken.
A lot of concepts in each video. Down the study hole I go!!! By the way, like many others have mentioned before, this course is great. Giving me the exposure to the latest DL curriculum that unfortunately my university lags behind in. So on behalf of the entire DL community, thanks a lot!
I feel this course is really great since this course is first one I have seen that binds the advanced theoretical content with practical implementation. Can't wait to get started. Thanks for the content!!
Amazing material! Pretty cool and intuitive visualizations, and love your energy on class! Code "tricks" are also great! Have been tipying "jupyter notebook" for years D: , until now! :D
Great lecture! Thank you so much! Is it possible to switch the big vs. small screens when appropriate? For example, when you write something on the whiteboard or try to explain something by gesture without the aid of animation or Twitter, make the classroom stream the big screen. Similarly, when you play the fabric animation or show the tweets, make the computer screen the big one. That would be fantastic!
I'm no longer using the whiteboard. And yes, I'm going full screen when I deem it necessary. They took 8 and 4 hours each to produce. I think I'm on the limit of how much time I can put behind each of them. Also, the camera is recording the entire wall behind me, so the square crop is already a zoomed version of myself.
Very good lectures ! May I suggest to the NYU technical team to work on the capture quality ? the sound is GSM like and saturated. The video resolution is very low. This would make this excellent course even better !
@@WilliamGacquer hahaha! It seems the entire 2021 edition is coming online as well. 😜 And it's fully remote, better quality, and redesigned and reorganised material.
Hi Alfredo, Thanks alot for the wonderful lectures, I didnt learn NNs this way, its amazing (Although, It made me think that I dont know any DL :p),. Most of the book/courses I read didnt explain the concept of transformation (affine, source and target space, like whats relation of bias with rotation) . Apart from this course and its website, Can you please share any reference book/text or course which introduces in the similar way as you did (Neural networks as general nonlinear transformers and how layer corresponds to these transformations)? I would be very thankful
I'm glad you like my lectures ❤️ It's okay, I believe I don't know much DL as well. Hmm, I haven't written a book about it just yet. This is stuff I came up mostly by my own, using some physics and intuition. I was planning on releasing a proper online course about some of the major intuitions. But again, no time and no money? Haha 😅 Maybe when I land my next job, if they pay me for this? Or I should put up a Patreon account.
@@alfcnz Thanks for the reply, yes. Completely agree with time and money constraint, The type of slides and intuition you bring will definitely require both. Hope we would see something in coming future, I am putting that in my watch list :D
Thanks Alfredo for making this beautiful course publicly available!!! What a great lecture it was.. I had never seen such a beautiful visual representation of what's happening under the hood... Just a quick question, do the notes provided on the website capture the whole idea of the lecture or do I need to make my separate notes for this course.....
This is so cool that you published this course for free, thanks a million. Question: are the weights initialised in a particular way to to the animation of the beginning (identity)? If not, how is it that the shape of the cloud is preserved until the last layer ?
@@alfcnz Oh yeah, I meant the animation at 15:00. If I understand correctly, what is ploted to the screen is from the two hidden nodes and with training the space wraps to make classification possible. So my question is, how do you make it so that at initialization, there is no transformation between the two input nodes and the hidden nodes. thanks :)
This course *does* teach PyTorch, and it does it from scratch. You're supposed to spend a few hour per notebook covered in the videos, after you've watched it. If you have questions, post them right here. If anything is broken, you can open an issue on GitHub (I reply quite promptly), or fix it yourself and send a PR (which would be highly appreciated).
I'm not able to reproduce the plot at 38:06 . I clone the Gihtub repo but my plot don't have red, Blue , green points. Can someone please help me with this !
I guess it’s basically the same as the notebook called 02-space_streching. The code is a little different than what appears in the video but it seems to cover the same material.
Hahaha! I had someone answering every question last year (Spring 2019) and I had to tell him “yes, I know you know the answer, just let others have fun too, okay?” hahaha! Answering my question because you're following is one thing, answering my questions because you already knew the answer is showing off 😛 Which one is you? 😏😏😏
Hahaha 😂 in this case just a vigorous follower😊, but what you told the guy in 2019, ie “...just let others have fun too”, reminded me of how I *sometimes* did not share cookies in kindergarten (oops), so I guess I’m a LINEAL COMBINATION of both 🤣 🍪😋
Why do you'll smart people keep giving lectures on basic material that we can already find so many tutorials on the web? Why don't you'll prepare accessible material for robust deep learning, adversarial training, bayesian deep learning, uncertainty estimation ...... such topics that actually require your expertise. We don't need PhDs and profs to tell us about how to use pytorch tensors.
Don't get me started about what is the average quality of entry level tutorials on the web. Moreover, this is the first out of 14 (?) lectures I teach my students and spent my weekends editing. Anyhow, you're free to hire me for teaching what you may need.
@@alfcnz I'm just saying I would like more advanced content from the few people who know and are willing to teach it well. How can I hire you? Is it on a per hour basis or what are the details here?
@@rahuldeora5815 I was being partially sarcastic. I teach what my students need to learn. Yann and myself spent quite some time planning the course in advance. Therefore, there's a reason behind the choice of content we showcase. I'm working on a new course with other faculty members for the next semester, but I'm only human. So, give me some time. Finally, yes, I teach individuals too (from middle school to my PhD students) with a tailored program and set of notions.
Hi, I am a student from a really poor country who really loves to learn about deep learning, and thanks to you it is now accessible to me. I am really grateful. thank you.
💪🏻💪🏻💪🏻
I hope that you understand the great value you're giving! I'm truly pleased watching your videos. You are giving us exposure for a lot of cutting edge technology/information/Concepts that is not available at our universities (or at-least when i was at University). Thank you for the hard work! If I can send you a picture it would be me with a paper and a pen, taking notes and studying your material.
Aww, that's adorable 😊
Yeah, I think I identified a gap in the current educational resources. Hopefully, I'll do justice to the task I've embarked and to the role I've taken.
So True😭
A lot of concepts in each video. Down the study hole I go!!!
By the way, like many others have mentioned before, this course is great. Giving me the exposure to the latest DL curriculum that unfortunately my university lags behind in. So on behalf of the entire DL community, thanks a lot!
On behalf of Yann and myself, you're most welcome! 😁
I feel this course is really great since this course is first one I have seen that binds the advanced theoretical content with practical implementation. Can't wait to get started. Thanks for the content!!
Yeah, we tried to be comprehensive and get you started with zero DL coding knowledge.
You're welcome 😉
Such a beautiful visual representation of what happens under the hood!!. Hats off Alfredo and Yann!
😊😊😊
this is the best intro i have ever heard from a teacher :D
Yay! 😊😊😊
Love it the way you explain and give hints visually. I never learned Neural networks this way ever.
I try to share my “view”. That's how I “see” things 🧐😋
Starred the repository so you got one more star... Loved your teaching style.... Wish I had found it much sooner and yes followed on twitter as well..
I'm glad you're enjoying it! I clearly am having fun, but it's great if you're having a good time as well!
i havent wacthed it yet ,..but i already feel like this is great and fun lecture from your enthusiasm..!!! ma ma mia =D
Hope you like it! Haha!
I appreciated the swirling graphic and went ooohhhh :)
Yay! 🥳🥳🥳
You are so interesting. I like your style of teaching. Thanks
Thank you! 😃 I try to entertain!
"This link will show you how to move stuff around!"
Thank you for this statement ^_^
Uuuuh… without time stamp I have no idea what link you're referring to 😅😅😅
@@alfcnz he is talking about 7:46 - link to Grant's video on linear transformations.
Amazing material! Pretty cool and intuitive visualizations, and love your energy on class!
Code "tricks" are also great! Have been tipying "jupyter notebook" for years D: , until now! :D
Thanks 🤗
Oh this looks good. I am so excited. A new series :D
Hehe 😁
I also like that you put the presentation digitally next to the presentation recording into another window.
@@st0ox getting myself and Yann to be confined in our boxes is the most painful part. I need to click a few thousands times per video.
Great lecture! Thank you so much! Is it possible to switch the big vs. small screens when appropriate? For example, when you write something on the whiteboard or try to explain something by gesture without the aid of animation or Twitter, make the classroom stream the big screen. Similarly, when you play the fabric animation or show the tweets, make the computer screen the big one. That would be fantastic!
I'm no longer using the whiteboard. And yes, I'm going full screen when I deem it necessary.
They took 8 and 4 hours each to produce. I think I'm on the limit of how much time I can put behind each of them.
Also, the camera is recording the entire wall behind me, so the square crop is already a zoomed version of myself.
Quanto vorrei tornare studente e seguire il tuo corso :D Thank you for making this available for free.
Puoi tornare a studiare quando vuoi. Il corso è qui per voi 😜
@@alfcnz siii quando riesco mi ritaglio del tempo per seguirti :D
@@liadiavoletto 🥳🥳🥳
Very good lectures !
May I suggest to the NYU technical team to work on the capture quality ? the sound is GSM like and saturated. The video resolution is very low. This would make this excellent course even better !
Hahahahahahaha!
There is no “technical team” behind this. It's all my work. Went a little crazy as well 😅😅😅
@@alfcnz I suggest renaming "NYU" to "NUY featuring A.Canziani & sponsored by A.Canziani" :)
Thank you very much, this content is a gem.
@@WilliamGacquer hahaha! It seems the entire 2021 edition is coming online as well. 😜 And it's fully remote, better quality, and redesigned and reorganised material.
I'm really happy..
And so am I 😊
Hi Alfredo, Thanks alot for the wonderful lectures, I didnt learn NNs this way, its amazing (Although, It made me think that I dont know any DL :p),. Most of the book/courses I read didnt explain the concept of transformation (affine, source and target space, like whats relation of bias with rotation) . Apart from this course and its website, Can you please share any reference book/text or course which introduces in the similar way as you did (Neural networks as general nonlinear transformers and how layer corresponds to these transformations)?
I would be very thankful
I'm glad you like my lectures ❤️
It's okay, I believe I don't know much DL as well.
Hmm, I haven't written a book about it just yet.
This is stuff I came up mostly by my own, using some physics and intuition.
I was planning on releasing a proper online course about some of the major intuitions. But again, no time and no money? Haha 😅
Maybe when I land my next job, if they pay me for this? Or I should put up a Patreon account.
@@alfcnz Thanks for the reply, yes. Completely agree with time and money constraint, The type of slides and intuition you bring will definitely require both. Hope we would see something in coming future, I am putting that in my watch list :D
Thanks Alfredo for making this beautiful course publicly available!!! What a great lecture it was.. I had never seen such a beautiful visual representation of what's happening under the hood... Just a quick question, do the notes provided on the website capture the whole idea of the lecture or do I need to make my separate notes for this course.....
Notes on the website were written by the students. Let me know if anything is missing.
I'm slowly writing a textbook, though.
This is so cool that you published this course for free, thanks a million. Question: are the weights initialised in a particular way to to the animation of the beginning (identity)? If not, how is it that the shape of the cloud is preserved until the last layer ?
You're welcome 😊
You need to add the minute:second of what you're referring to, otherwise I cannot understand your question.
@@alfcnz Oh yeah, I meant the animation at 15:00. If I understand correctly, what is ploted to the screen is from the two hidden nodes and with training the space wraps to make classification possible. So my question is, how do you make it so that at initialization, there is no transformation between the two input nodes and the hidden nodes. thanks :)
Sei un grande!
Grazie! 😎😎😎
Great lecture series 👏👌
Thank you ❤️
Simply super!
😏😏😏
Great course!
Thanks!
thank you
Welcome!
1_000 == 1000
I have been coding in python way too long to have not known that. How is that use not more common?
Not sure. I get confused if I see too many digits 😅😅😅
eccellente, bellisimo
Grazie 🥰🥰🥰
I haven't learned pytorch. Should it be a good idea going along with the course and learning pytorch? I know basic python and numpy though.
This course *does* teach PyTorch, and it does it from scratch. You're supposed to spend a few hour per notebook covered in the videos, after you've watched it.
If you have questions, post them right here.
If anything is broken, you can open an issue on GitHub (I reply quite promptly), or fix it yourself and send a PR (which would be highly appreciated).
@@alfcnz I would love to contribute anyway.
Thanks for the quick reply. It's motivating.
I'm not able to reproduce the plot at 38:06 . I clone the Gihtub repo but my plot don't have red, Blue , green points. Can someone please help me with this !
Yes, I haven't pushed the multi colour version. It's in my to-do list. (And it's been there for a year at least.)
good content
🙏🏻
Where can I find the Random Projections notebook shown around minute 30? I couldn’t find it in the GitHub section pytorch-Deep-Learning
I guess it’s basically the same as the notebook called 02-space_streching. The code is a little different than what appears in the video but it seems to cover the same material.
@@ikejimenez3836 yeah, I should push the newer version. I need an assistant, hahaha 😅
hihihih can't see the whiteboard :'(
This year everything is remote and I'm drawing on a iPad. I've also started using an iPad last year after noticing the unreadablility issue.
you are a funny prof
Yay! 😋
Haha would like to be there and answer effusively each question **nerd**
Hahaha! I had someone answering every question last year (Spring 2019) and I had to tell him “yes, I know you know the answer, just let others have fun too, okay?” hahaha! Answering my question because you're following is one thing, answering my questions because you already knew the answer is showing off 😛
Which one is you? 😏😏😏
Hahaha 😂 in this case just a vigorous follower😊, but what you told the guy in 2019, ie “...just let others have fun too”, reminded me of how I *sometimes* did not share cookies in kindergarten (oops), so I guess I’m a LINEAL COMBINATION of both 🤣 🍪😋
@@gastonmazzei8087 🤣🤣🤣
audio is bad
Watch the new edition? 🤷🏼♂️🤷🏼♂️🤷🏼♂️
Why do you'll smart people keep giving lectures on basic material that we can already find so many tutorials on the web? Why don't you'll prepare accessible material for robust deep learning, adversarial training, bayesian deep learning, uncertainty estimation ...... such topics that actually require your expertise. We don't need PhDs and profs to tell us about how to use pytorch tensors.
Don't get me started about what is the average quality of entry level tutorials on the web. Moreover, this is the first out of 14 (?) lectures I teach my students and spent my weekends editing.
Anyhow, you're free to hire me for teaching what you may need.
@@alfcnz I'm just saying I would like more advanced content from the few people who know and are willing to teach it well. How can I hire you? Is it on a per hour basis or what are the details here?
Because his way to explain it is so good ^^
@@rahuldeora5815 I was being partially sarcastic. I teach what my students need to learn. Yann and myself spent quite some time planning the course in advance. Therefore, there's a reason behind the choice of content we showcase.
I'm working on a new course with other faculty members for the next semester, but I'm only human. So, give me some time.
Finally, yes, I teach individuals too (from middle school to my PhD students) with a tailored program and set of notions.
There are tons of papers on arxiv etc., that you can get as deep as you want