- 10
- 29 520
Great Fate
Oman
Приєднався 1 бер 2016
I just post stuff I think would be interesting for you guys.
The channel banner is generated using DALL-E-2 :D
The channel banner is generated using DALL-E-2 :D
What Are Neural Networks Even Doing? (Manifold Hypothesis)
In this video, I try to crack open the black box we call a #neuralnetwork 💪💪
The animations were made using #Manim Community Edition, and here is the code I wrote to generate them:
github.com/igreat/videos/tree/main/manifold-hypothesis
Timestamps:
0:00 - recap
0:49 - visualizing neural networks 2d
3:05 - linear transformations
3:27 - nonlinear transformations
4:30 - affine transformations
5:14 - back to 2d neural networks
7:04 - why use more neurons per layer?
10:10 - manifold hypothesis
11:36 - visualizing handwritten digit separation
12:56 - conclusion
The animations were made using #Manim Community Edition, and here is the code I wrote to generate them:
github.com/igreat/videos/tree/main/manifold-hypothesis
Timestamps:
0:00 - recap
0:49 - visualizing neural networks 2d
3:05 - linear transformations
3:27 - nonlinear transformations
4:30 - affine transformations
5:14 - back to 2d neural networks
7:04 - why use more neurons per layer?
10:10 - manifold hypothesis
11:36 - visualizing handwritten digit separation
12:56 - conclusion
Переглядів: 15 049
Відео
How Do Computers See Images? - Neural Networks
Переглядів 8 тис.2 роки тому
Hello everyone, this is a video that I made in a little over a week and I had originally planned to have all 3 parts into one. However, due to time constraints on #SoME2, I had to push most of the exciting and "less talked about" stuff into part 2 and 3. Note that this mainly serves as a refresher on neural networks to lead to part 2 and 3, but I think it could also work quite well as a first i...
Think Twice Before You Build A Time Machine - short animation
Переглядів 2032 роки тому
This is my final art project. I actually had very little motivation to finish it which is why I made it scuffed.
Fun Fractal Project
Переглядів 4372 роки тому
Sierpiński triangle. Simple triangle fractal. Think of every iteration as multiplying the area by 3/4. If I multiplied infinitely many times, the area approaches 0!
How I Went From 50wpm to 100wpm in 2 Months.
Переглядів 5 тис.4 роки тому
Note: for the vast majority of the days, I only did 10-20 minutes of practice, something that is sustainable yet also just enough to get you improving. This is the story of my typing journey, and how I went from 50wpm to 100wpm in just 2 months... anyways, this was my first time on sony vegas for years so I apologize for the scuffed editing. Plus, this is also the first time I make a video in t...
سيف النار - موات
Переглядів 1644 роки тому
hello, this edit was originally just me practicing the tracking tool, but I soon decided to make it into a full-on project, but like I said this was originally just practice and so I did not put as much effort into it as I would've with a regular project. Nonetheless, I am proud of the outcome :).
Halfway love - Kimi no Na wa: AMV
Переглядів 1464 роки тому
What's up boys, this is the first AMV I've ever made so it might not be as smooth as I wished it to be. But overall I'm fairly proud of the outcome :)
Great animations
Wow! This is so intuitive! Thank you so much 👌
incredible video!
Still looking for that source on relu in the description :P
Gem
Fantastic video
Beautiful animations.
Insanely cool animations. really helps to visualize.
hey this video is insane, which softwares you have used to make these
Really great, thank you
Can you please continue making videos, this is a gold mine! much love.
Very clear explanation, thank you. I would appreciate it if you could delve into conv nets and how the convolution/cross correlation is converted to a matrix operation. I find it difficult to understand and everyone just tells me to use pytorch and not think about it...
Why did you stop posting?
heck yeah
Awesome demo of neural network. Can you tell me what plotting tool you used fir the last 3 dimensional plot showing clusters of similar numbers? Thanks.
This is surprisingly deep and informative! Quite rare on YT these days. Great job!
Reddit says this isn't how neural networks work lol
wow. at the start of my journey within deep learning these insightful and intuitive tutorials didnt exist. this is everything wrapped up neatly and connected. very very nice
How the hell do you only have 706 subscribers? This is some of the best visualization I've seen of NNs on any channel.
just WOW! a totally new perspective
amazing video bro, one of the best on NNs. Trust me I have seen many and this one stands out. Sadly this is not famous, shocking
karpathy's convnetjs site has some 2d demos similar to this with manifolds, this vid makes it way easier to understand whats going on
I'd love to learn more from you. It's weird how much irrelevant stuff one as to filter through to get to gold nuggets such as yourself. UA-cam seems more interested in giving me stuff I didn't ask for in searching
Thanks mate, this was interesting! The universe is beautiful! ❤
Wow! Such a fascinating video, beautifully animated😀
You don't know it, bro, but you are a gem. I love your content.🙏
عمل رائع ومشرف بارك الله فيك
As the others commented already, this is of unique quality and clarity of exposition. Take heart and have faith that your continued work will find greater appreciation!
Cool video! Please do more :-)
Very nice.
why is a video this amazing, have so little views? I don’t understand!
Best intuition guide for NNs I've seen yet
I love your videos. It is so useful!!
Really nice video man. Keep on the good work
cringe
Crazy good video.
Your volume is too low
My boi didn't put other resources in the description as promised :(
So awesome!!!
This is a fantastic video, and the visualizations are great as well. I’m interested to see how you visualize convolutions in your next video!
Thanks! I appreciate it coming from someone who makes wonderful visuals themselves
Excellent video, well done!
Nice!
Awesome Visualization! Never thought of NNs that way :)
After 20 years in UA-cam, i think this might be the first UA-cam comment I've ever written. Excellent video. Thank you. Neutral networks are obviously becoming very important. I appreciate having a slightly better technical understanding of what is happening inside that magical black box. This is the first video that has really been able to do that for me. I subscribed, and look forward to seeing more from you.
Thanks a lot this comment makes me so happy 🙏🙏❤️
Great video and amazing work thanks
Wow! much quality! really good explanation and visuals I finally get what NNs does now
This literally blew my mind away✨
This is a great way to visualize and explain the layers. Thank you for the effort you put in to this video!
Very nicely done, I've seen these concepts before but the animations do a much better job at getting across the process than static images do. I'd love to see the intuitive examinations of lstm and transformer (and looking forward to the CNN one you're working on!) Subbed!
Thanks a lot for the kind words 🙏 I explored LSTMs and Transformers deeply a while ago and certainly had the idea to make a video exploring them (maybe my next videos after CNNs? 👀👀)
@@greatfate I would love to see the process of all of these architectures during training - the attention matrices in a transformer, the cell state of LSTM, the kernels of CNN etc. I really enjoyed that you used PCA to help visualize that process for the vanilla networks.
One thing though....change the screen background to dark! Dark themes are one the rise!
Hmmm… frankly I specifically chose a light themed color palette to differentiate myself from other similar content. But now that you mentioned it… I’ll experiment with dark themed color palettes 🎨 and see if it’s worth the switch 👊
@@greatfate Don't get me wrong I'm all for the white background, but half the time I'm watching at night. Maybe a darker grey?