But what is a Neural Network? | Deep learning, chapter 1 

Skatījumi 9 milj.
99% 86 719 616

Home page: www.3blue1brown.com/
Brought to you by you: 3b1b.co/nn1-thanks
Additional funding provided by Amplify Partners
Full playlist: 3b1b.co/neural-networks
Typo correction: At 14 minutes 45 seconds, the last index on the bias vector is n, when it's supposed to in fact be a k. Thanks for the sharp eyes that caught that!
For those who want to learn more, I highly recommend the book by Michael Nielsen introducing neural networks and deep learning: goo.gl/Zmczdy
There are two neat things about this book. First, it's available for free, so consider joining me in making a donation Nielsen's way if you get something out of it. And second, it's centered around walking through some code and data which you can download yourself, and which covers the same example that I introduce in this video. Yay for active learning!
I also highly recommend Chris Olah's blog: colah.github.io/
For more videos, Welch Labs also has some great series on machine learning:
For those of you looking to go *even* deeper, check out the text "Deep Learning" by Goodfellow, Bengio, and Courville.
Also, the publication Distill is just utterly beautiful: distill.pub/
Lion photo by Kevin Pluck
0:00 - Introduction example
1:07 - Series preview
2:42 - What are neurons?
3:35 - Introducing layers
5:31 - Why layers?
8:38 - Edge detection example
11:34 - Counting weights and biases
12:30 - How learning relates
13:26 - Notation and linear algebra
15:17 - Recap
16:27 - Some final words
17:03 - ReLU vs Sigmoid
Animations largely made using manim, a scrappy open source python library. github.com/3b1b/manim
If you want to check it out, I feel compelled to warn you that it's not the most well-documented tool, and has many other quirks you might expect in a library someone wrote with only their own use in mind.
Music by Vincent Rubinetti.
Download the music on Bandcamp:
Stream the music on Spotify:
If you want to contribute translated subtitles or to help review those that have already been made by others and need approval, you can click the gear icon in the video and go to subtitles/cc, then "add subtitles/cc". I really appreciate those who do this, as it helps make the lessons accessible to more people.
3blue1brown is a channel about animating math, in all senses of the word animate. And you know the drill with LVlocal, if you want to stay posted on new videos, subscribe, and click the bell to receive notifications (if you're into that).
If you are new to this channel and want to see more, a good place to start is this playlist: 3b1b.co/recommended
Various social media stuffs:
Website: www.3blue1brown.com
Twitter: 3Blue1Brown
Patreon: patreon.com/3blue1brown
Facebook: 3blue1brown
Reddit: www.reddit.com/r/3Blue1Brown

Publicēšanas datums


5 okt 2017






Mans atskaņošanas saraksts
Skatīties vēlāk
komentāri 100   
Ren chon
Ren chon Pirms 12 Stundām
At first I thought there was a typo in the code at 15:15, and then I was like, oh shit, this is genius
Daniel Tang
Daniel Tang Pirms dienas
8:18 r e c o g n i t i o n --> re cognition omg how have I never seen this
Aamna Suhail
Aamna Suhail Pirms dienas
BuckWheat Mycelium
BuckWheat Mycelium Pirms dienas
So is the structure of a neural network essentially a large hashing function?
Anthony Maylath
Anthony Maylath Pirms dienas
Great job with the explanation in this video! 9M views is really challenging to obtain for such a technical topic. Well done!
Septianti Dwi Purwasih
Septianti Dwi Purwasih Pirms 2 dienām
Dharma 111
Dharma 111 Pirms 2 dienām
how is your voice so perfect compared to anyone else who speaks? Everyone sounds like shit next to you
plekkchand Pirms 3 dienām
This is a superb expounder on things mathematical. He introduces Multivariable Calculus on Khan Academy as well, (where he achieves the remarkable feat of introducing every single video with the word "So"- but I cavil).
Onyinyechi Chukwuma
Onyinyechi Chukwuma Pirms 3 dienām
Wonderfully presented! Thank you very much :)
Stavros Ach
Stavros Ach Pirms 3 dienām
If you need money buy BTC.if you cant buy start mining, BTC 100.000$ in 2021.Ask me how to mine BTC.
Johnny Johnny
Johnny Johnny Pirms 4 dienām
I'm sorry, what?
Ben Son
Ben Son Pirms 5 dienām
So i suck at math while my brain does matrix and adjebra just so i can understand my math tacher's ugly handwriting!! Brain! this is not fair!!
Tanya Pirms 5 dienām
Wonderful video :O
Samvardhan Vishnoi
Samvardhan Vishnoi Pirms 5 dienām
squishification function
Jahed Hossain
Jahed Hossain Pirms 6 dienām
omg! You are so good and amazing and so kind to make such amazing lecture materials!
Atharva Garole
Atharva Garole Pirms 6 dienām
You know what's better than a neural network? Just take a child and then make him do all the work for free
Markus Österle
Markus Österle Pirms 7 dienām
I am new to this topic and I always wondered if it is necessary to have 10 output nodes. Couldnt you have 4 output nodes and interpret them in binary and train the network to output the number in binary as well? I dont know if this would make anything simpler or more complicated I was just wondering if that is in principle possible. Would be useful for something with a lot of output nodes I would imagine.
klim po
klim po Pirms 7 dienām
it's my 5th time watching this video. Let's hope I understand it this time.
better world
better world Pirms 7 dienām
Finally learned something useful about NN. Great work. The quality is unbelievable. So understandable.
حيدر كريم
حيدر كريم Pirms 8 dienām
Thank you so much 🙏🏻
FOR Vpbof-20
FOR Vpbof-20 Pirms 9 dienām
Thank you 3B1B. I have watched all 4 lectures and re-watched them backward practicing what I learned in back propagation. Hey if it works for neural network it should work for me too.
Maitha Khanji
Maitha Khanji Pirms 9 dienām
very good and clearly explained video with amazing graphics, thanks a lot! link for video was provided by professor to understand neural networks
Çağrı Kartal
Çağrı Kartal Pirms 9 dienām
ThomasDrijusl14 Pirms 9 dienām
I don‘t really understand why the weights are just random numbers. Doesn‘t it make a difference for the weighted sum whether the weight is e.g. 1.76 or 3.41? What determines the numbers of each weights? Or is it just important if the weight is positive or negative?
Zeba Shaikh
Zeba Shaikh Pirms 4 dienām
A weight decides how much influence the input will have on the output. it controls the signal between two neurons. weights later then gets multiplied by activation function then if activation fun is high it will fire neuron else not. means that output having more value will popup. I hope I helped else you should watch it again and also research more about weight and biases in neural network.
marina anand
marina anand Pirms 10 dienām
Great explanation. Cheers.
Chuckk Hubbard
Chuckk Hubbard Pirms 10 dienām
A small beef - neural networks aren't inspired by the brain; they're in the brain. The ones in a computer are more "neuron-y" like that "chocolatey" cereal or "cheesey" crackers that don't actually have chocolate or cheese.
Liam Waddle
Liam Waddle Pirms 11 dienām
Easily the most beautifully and clearly presented video on computer science I've seen. Amazing.
P P Pirms 11 dienām
ak 47 tec 9
Mike Orr
Mike Orr Pirms 11 dienām
I actually made a small version of this in MineCraft years ago on a 5 by 5 grid. You draw a number 0-9 onto a 5x5 grid and it guesses the number you drew. Worked pretty good and was a lot smaller than any MC calculator.
kryptoid 256
kryptoid 256 Pirms 11 dienām
self in python in java,c,c++,c#,etc., we call it the this pointer is the adress of whatever what object to be manipulated by the function
Syeda A Hamdani Official
Syeda A Hamdani Official Pirms 12 dienām
It's soo much difficult to trained a model look God has created our brain with soo much perfection
labros labrou
labros labrou Pirms 13 dienām
nruh, you didnt give any kind of example.
Sssrnty Pirms 14 dienām
Thank you for this!
Yahya Khan
Yahya Khan Pirms 15 dienām
You should paste the chapter 2 ,3 4 or related to the topic link in the bio ..just a suggetsion!
RESHMA PATIL Pirms 15 dienām
Tqs !! 🤩
Abhiram Pirms 16 dienām
Every second of this video is a Pre-requisite to the next second of the video :D
Denis Kazakov
Denis Kazakov Pirms 16 dienām
Remarkably good explanation!
Loïc Dallaire
Loïc Dallaire Pirms 16 dienām
emirhan Pirms 16 dienām
What should I learn to dive into these topics
Δημήτρης Βουζουναράς
Why we like to make the things more complicated? The sigmoid can be rewritten as σ(x) = e^x / e^x + 1. But why we choose (e) and not (2) as base, that's a lot easer for binary computing?
Z Fang
Z Fang Pirms 17 dienām
14:40 the bias vector should be k by 1?
Just For Fun
Just For Fun Pirms 17 dienām
Thank You !!!
Just For Fun
Just For Fun Pirms 17 dienām
I Love this Channel !!!
SaltTheory Pirms 17 dienām
What an absolute legend. Thank you for explaining this!!!
Modernong Deista
Modernong Deista Pirms 17 dienām
This is the only time that somehow I understand how neural network works. Thanks for speaking human @3Blue1Brown. They say AI is not automation (computer programming). So "Artificial Intelligence" is basically just automation as well but using advanced programming, mathematics and complex algorithm. Then I can say it's creator, the human with organic brain is much much more advanced. hehe
Prakhar sharma
Prakhar sharma Pirms 18 dienām
A lot of this spent on videography Apriceat your work sir
Daniel Lamperouge
Daniel Lamperouge Pirms 19 dienām
So satisfying to see these animations. I'd love to put animations like this into my presentations at uni… but let's first get me through my Neural Networks exam :D
Alison Murphy
Alison Murphy Pirms 19 dienām
The thin piccolo psychologically stitch because slash contrarily trap amid a abject pot. silent, clever cheek
iCraftC4 Pirms 19 dienām
3:37 I love this part so much, i didnt understand neural network layers at all before this that one little visual helped a lot
Abdo Mohamed
Abdo Mohamed Pirms 19 dienām
Epic ...
A little meow
A little meow Pirms 19 dienām
late to the party.
shasherazi Pirms 20 dienām
how do i like more than once?
Tejas G. Shende
Tejas G. Shende Pirms 21 dienas
Yet another fine video by 3B1B - much appreciated! Btw I had a small query regarding the ReLU function mentioned at the end - if ReLU = max(0,a) then how do we confine it to the range [0,1] for any neural cell to accept the value? I mean, we started out with the sigmoid function for the very reason that it "squished" the number line into the interval [0,1] so how is ReLU being used if it doesn't even serve the basic purpose it was meant to serve?
Sergio Molina
Sergio Molina Pirms 21 dienas
skierpage Pirms 21 dienas
But ReLU() doesn't normalize to 0..1 as sigmoid does. Its max value in the first inner layer is the number of inputs (784.0). With ReLU are the inputs treated as int or float by the next layer? You can see why specialized chips and numerical representations matter so much in AI implementations.
Juan Hernandez
Juan Hernandez Pirms 22 dienām
The chemical gas chemically play because moustache endoscopically intend following a chilly bengal. warm, acoustic cast
Guilherme Daniel
Guilherme Daniel Pirms 22 dienām
A neural network to understand doctor letters.
idegteke Pirms 22 dienām
Wait, wait, wait! Stamping our judgement (of e.g. which picture is which number) into a neural network can, in theory, be MAXIMUM as “intelligent” as the resource that serves as a reference (in the teaching phase). It’s all about giving the best possible estimation of a result that the reference (person) would come up with. Okay, it will, eventually, be more precise and vastly faster in using the imprint. This imprint (facial recognition or a certain level of autonomous physical situational awareness of a dynamic robot, or machine translation for example) is sometimes misinterpreted as “intelligence” while it’s anything but. HOW this rubber stamp is supposed to turn into something that is BEYOND stamping and stamping the same imprint mechanically?
Serban Adinan Alexandru
Serban Adinan Alexandru Pirms 22 dienām
is there a way to feed raw footage of videos to the neural network that they can replicate in a game?
viswanathan sreyas
viswanathan sreyas Pirms 23 dienām
Genuinely awesome. Thank you 3blue1brown!
Sumatwawa Pirms 23 dienām
What do you think about R ? Worse than python (especially concerning deep learning, bc good libraries also exist : keras, tensorflow, caret ...) Any advice please ?
Emily Jackson
Emily Jackson Pirms 23 dienām
The innocent street biophysically grease because hawk cumulatively soothe into a nice keyboarding. , icky dish
Meneer Leonard De Hertog van Cydonia
Some believe that artificial intelligence are thinking, but is not so. It is just following some mathematics, computer codes and logic rules.
عدنان هشام محمد
This system is just space!Cоme see me to watch the lаtest vide0.
vansh jhunjhunwala
vansh jhunjhunwala Pirms 26 dienām
The nueral network in my mind lit up and shouted "linear algebra" when you said "let me show you a compact notation." At this point, there is no escaping Linear algebra, it's everywhere
lexxsimf Pirms 26 dienām
I am watching how programmer solving task of recognizing letters, that was solved more than two decades ago. Again, jus new terms to well written programs long ago.
hackey mabel
hackey mabel Pirms 27 dienām
cảm ơn và xin lỗi bạn nào dịch tiếng Việt nha qua 3 lần xem lại khúc chuyển qua dùng khối lượng mình cũng chưa thông đc T..T
Adam Doughty
Adam Doughty Pirms 28 dienām
Loofus Doofus
Loofus Doofus Pirms 28 dienām
This is A Super Educated Video And Will Inspire More And More People :D (btw I edited This 2 times :) Ok!)
Quade Pirms 28 dienām
I can't believe he didn't animate all 12960 connections. How rude - he only animated 672!
Mark Robert Piazza
Mark Robert Piazza Pirms 28 dienām
I am wanting to learn how to present information in this type of format (i.e., graphical animation style). I am very unexperienced when it comes to graphic design and video editing but I want learn ways to present technical information in a more visual and intuitive way. It would be greatly appreciated if anyone could point me in the direction of resources related to how one would go about creating this style of video. Thanks!
Arun Bharadwaj
Arun Bharadwaj Pirms 28 dienām
Thank you so much..
hackey mabel
hackey mabel Pirms 29 dienām
how can you regconise the number 3? Well, my teachers are sucks at writing it correctly
Larson Blanca
Larson Blanca Pirms 29 dienām
When someone loves you, the way they talk about you is different. You feel safe and comfortable.
435lov Pirms Mēneša
This is too good!
Gregorio Tonelli
Gregorio Tonelli Pirms Mēneša
Ryan La
Ryan La Pirms Mēneša
no you get destroyed marcos like what the hecke
Ryan La
Ryan La Pirms Mēneša
you dont know how to speel
Leo Müller
Leo Müller Pirms Mēneša
I think at 14:51 the bias vector must be from R^k because the product of the weight matrix and the activation vector is a vector of dimension kx1. Really well explained! Thanks!
A Arjith
A Arjith Pirms Mēneša
Are there only 4 episodes?
Wassan Saad
Wassan Saad Pirms Mēneša
Hi, Can I know what is the tools or applications you used to make these wondeful slides. I need to do luctures to my students in the same way. Many thanks in advance
AskRedditReader Pirms Mēneša
The main purpose of the sigmoid is not to squash the values between a certain interval 0 - 1 as has been said here. Its main purpose is to introduce non linearity into the network.
agentdark64 Pirms Mēneša
Will this work against recapture? and identifying car images?
Lora Bowers
Lora Bowers Pirms Mēneša
Yang pilih Allah like yang pilih Dajjal abaikan.
Oliver Schneider
Oliver Schneider Pirms Mēneša
Hey, you there. Have you read the corresponding book 3blue1brown linked below this video? I watched the series 2 years ago, and I am unsure if I should watch the video series again before reading the book or after. Thanks, have a nice day.
Villuri Kishore
Villuri Kishore Pirms Mēneša
I will forever be grateful to you for making learning so much fun!
Matthew TIKI
Matthew TIKI Pirms Mēneša
This is gold, thank you so much!
Abdulğani Yıldırım
Abdulğani Yıldırım Pirms Mēneša
Everything in the human brain is incredible and very very detailly designed. It cannot be...
Muhammad Miradeljan
Muhammad Miradeljan Pirms Mēneša
May we have this video?
Riddhi Manna
Riddhi Manna Pirms Mēneša
It felt like I was watching an anime didn't even feel like I was putting in an effort and I learnt so much! Thanks Grant!
Jakub _
Jakub _ Pirms Mēneša
Great video!💓
Alexander Medina
Alexander Medina Pirms Mēneša
The last time I had the feeling I got watching this video was when I first understood derivatives in my fresh year at university. Thank you for this.
hy q
hy q Pirms Mēneša
this is awesome!
Narayana Babji
Narayana Babji Pirms Mēneša
OCR. .
Narayana Babji
Narayana Babji Pirms Mēneša
If the input 3 is flipped vertically It gives a output of 5. wrong output and a anomaly. .
Tenzin Passang
Tenzin Passang Pirms Mēneša
I'm only here because of Elon Musk and to make sure that neural net is not some kind of a fishing net. Well, he was right. lol
Michael Mccoll
Michael Mccoll Pirms Mēneša
can it "know" if its not a nuber
Hend Muhammad
Hend Muhammad Pirms Mēneša
Bilal Pirms Mēneša
"Before plugging it through the sigmoid squishification function"
Paulo Priolli
Paulo Priolli Pirms Mēneša
My head....
Lucas Parmentier
Lucas Parmentier Pirms Mēneša
Very well explained !
Осмотр BMW 530e iPerformance
12a: Neural Nets
Skatījumi 384 tūkst.
The Essence of Calculus, Chapter 1
AlphaGo - The Movie | Full Documentary
Осмотр BMW 530e iPerformance
Skatījumi 8 milj.
UFC 259: Post-fight Press Conference