How Deep Neural Networks Work - Full Course for Beginners
HTML-код
- Опубликовано: 28 апр 2024
- Even if you are completely new to neural networks, this course will get you comfortable with the concepts and math behind them.
Neural networks are at the core of what we are calling Artificial Intelligence today. They can seem impenetrable, even mystical, if you are trying to understand them for the first time, but they don't have to.
⭐️ Contents ⭐️
⌨️ (0:00:00) How neural networks work
⌨️ (0:24:13) What neural networks can learn and how they learn it
⌨️ (0:51:37) How convolutional neural networks (CNNs) work
⌨️ (1:16:55) How recurrent neural networks (RNNs) and long-short-term memory (LSTM) work
⌨️ (1:42:49) Deep learning demystified
⌨️ (2:03:33) Getting closer to human intelligence through robotics
⌨️ (2:49:18) How CNNs work, in depth
🎥 Lectures by Brandon Rohrer. Check out his RUclips channel: / brandonrohrer
🔗 Find more courses from Brandon at end-to-end-machine-learning.t...
--
Learn to code for free and get a developer job: www.freecodecamp.org
Read hundreds of articles on programming: medium.freecodecamp.org
And subscribe for new videos on technology: ruclips.net/user/subscription_cent...
I just woke up. I am very confused. Why am i here-
Same lol.
It’s an exclusive club we’re learning in our sleep
Like seriously same just happened 😂😂😂
🌾👀🌾
Same 😂
Just woke up, don’t know where I am or how I ended up here
same, last thing i remember was veritasium explaining game theory
Bro same i was watching fresh spawns i think...
Ouch.. same here
Same, was watching serpentza.
same😭
So, I just woke up to this video on my phone but the ironic part is I just learned about this yesterday.
Same as me and I kept the link in a file for learn it next time.
me too😂
Oh and I didn't learn about this and have 0 interest in this...
Same here
Same. We are all connected.❤
Let me guess: you just woke up and this video was playing
and it’s already an hour and thirty three mins in, how tf?!?!??
YESSS
HOW DO YOU KNOW
YES
woke up 44 minutes in
dunno why but this video was playing when i woke up in the middle of the night
i slept watching a different completely unrelated video and woke up on this what just happened
Yup
Same!
Singularity trying to nudge you in the right direction
You changed the weight without adjusting for the error. Happens all the time
this night same happned with me
Watching this on my way to sleep for all the people who are waking up to this, it might break the cycle. 🙏💪
It didn’t 😢
Somehow this autoplayed on my phone while I was sleeping.
Realy?
I came with an interest in neutral networks.
I left feeling well rested.
I fell asleep watching a very simple maths video and woke up to this after dreaming that me and my friends were studying its contents. I’ve never done anything to do with this before but I understood it when I was dreaming about it so will probably give it another listen. It reminds me of being in College/University when SWIM was doing a bunch of drugs and accidentally designed a computer brain. Score for drugs 1,264,273,995,267,177, score for sobriety: still zero LOL
SWIM... There was an online forum I used to frequent. It's been years. I don't recall how to get there. I assume you know which I'm talking about. Does it still exist?
YAY drugs
@@therainbowtrout1820yes albeit not necessarily in the same regard
@therainbowtrout1820 It could be Bluelight which is popular. There was another one I used many years ago, but I can't remember the name of it. I'm not sure it exists anymore.
From sleeping on a Geopolitics video to landing here, I am stunned😅
So to be the first, I’d just like to say my journey consisted of falling asleep to a video about why a magnet on the front of a car wouldn’t work, then it went to cursed units of measurement, then it went to professor Dave explains and then I ended up here, all in all I’ve been asleep for about 3 hours and I need more sleep…
Anyone else wanna share the journey?
I started by watching “why therapy sucks for men” I then fell asleep, and RUclips showed me what gaming does to my head, to then finish here, it would’ve been Waaaay more if didn’t have my console on auto rest mode
I woke up and this was playing on the background
Seriously. Same here
@@MossawirAhmed
Bruhhh same tf
🤣🤣🤣
Just woke to this playing. It was the catalyst to the craziest most vivid dream since childhood...im in my 30's.
RUclips is a good detector of sleep
Oh, so I wasn't the single one falling asleep watching something then ended up here being confused
Amen
@@marius.y6360 me as well😂
@@marius.y6360 ⁹
Ur right
Very surprised to see everyone woke up to this video as well. The algorithm strikes again!
Apparently everyone waking up to this including myself 😂
is this a joke or truth? what exactly happen?
@@fatemehmohseni5414 load of people including myself suddenly wake up to this video. Autoplay at it's finest
Yes I'm scared too
It’s making me laugh so hard 😂
what is this, i just woke up..
I watched this on purpose. :) Found it quite helpful! Cheers
Assuming that everyone has had or currently has a learning capacity you realize that environment plays a huge part.
I’m pretty happy to be awaken by a such interesting lecture. Will watch it again
True! I woke up after it ended and the headline was interesting enough to hit replay while awake 😂
If you're an audio guy, Squash functions are just compression by factor of Ratio (r). Threshold is the pickup weight input and knee is smoothing of weights between input and output over a certain range. And there you go. Compression in a nutshell. However the dB peak scale is non-linear. The dB scale is power of 2x10dB. That's what makes it the most confusing.
So a ratio of 10 to keep it simpler is double the volume at the threshold gradually weighting less until the set peak where compression is zero. The knee rolls off that effect by a dB factor at a specified loudness and breadth of its impact. Seems gaussian to me. I don't know how the math works at the knee but it gives a smoother transition from boosted to left alone. So in a typical simple compression threshold at -24dB with 10 ratio would result in threshold at -12dB tapering to -10dB, -8dB, -6 and so on until you hit zero assuming your highest peaks are 0dB which is bad. Then you adjust the output to -8 or -14 depending on the sound and that scales the whole curve downward unaltered relatively by whatever output dB you set. If your threshold was boosted by compression to -12dB and you scale it down in output by -8dB then your threshold after processing will be -20dB tapering off up to -8dB in the same curve it had before the output was scaled down.
That's why you have to adjust input vs threshold vs ratio vs knee vs output to get the best out of simple compression. Multiband compression is the same thing just much more complicated as it accounts for frequency where you can specify within a certain frequency range how much compression you'd like. Overlap them and yeah that gets quite complicated but it's super useful to getting the right sound especially in dialogue to grab and manipulate the loudness of tonality and sibilance while rejecting the background noise or any echo or unwanted reverb.
The same principles apply in NNs in more of a deterministic and mathematical way. It entirely depends on the architecture and what it is used for as you are taking a larger dynamic range of inputs and compressing them to a smaller range of outputs. That's why CDs in the 90's Redbook audio was 16 bits wide. 2^16 made for 65536 levels of volume for any given sample. That was enough because it was replacing cassette tape which had horrible dynamic range. Now it's standard to have 24 bit audio which has a vastly higher dynamic range of 16,777,216 levels of volume at any given sample. For production and processing it's common to have 96 bit audio which has 7.92281625 x 10^28 levels of loudness. That's technically not better than analog but no human would ever be able to tell the difference. It helps computers and audio processing make very very accurate changes.
oh word !
Quite a pointless post really. Going to tremendous depth using an analogy to explain neural networks.
Far better to understand the network rather than your analogy. And yes, I used to work in audio engineering.
Analogies are useful as a means of explaining, of education, but your analogy is so specialised it has very little use in educating people.
Excellent explanation, excellent figures, and animations, awesome speaking! Looks like a dream course!
I suddenly opened my eyes and dreamed about this video while sleeping with my tai chi instructor at my beachfront property. Unreal.
Wow first time I’m actually glad I learned calculus in school. Nice to see it useful outside of the classroom.
T
भघ
Yep. Everyone thinks they know better. "I'll never use this!" Then why are they trying to teach it to you? 🙄
just cus they tryna teach it doesnt mean its necessary or objectively useful. most schools dont teach how to do taxes, and those are mandatory @@zilog1 🙄
@@zilog1 Most people never use it again
Thanks 🙏 I finally have some understanding of why cnn’s work!
Thank you for making this!
I knew an engineer brother of a friend who was working on how best to implement gradient descent into NNs years and years ago. I think he was one of the ones who gave up before CNNs became a widely used method. He certainly isn't a NN engineer anymore. He went on to predictive logistics which resembles RNN but really it was a much simpler feedback loop and balancing input versus output. Part of the Just in Time production to delivery process. Likely, the processing power and tech in the 90's wasn't powerful enough to realize the emergence big data is capable of now. Kinda wonder what he would have done had he been doing that 25 years later than he was. I know that he uses advanced NNs now of various types for his job but at this point he is an implementer rather than a developer. Tuning plays a big role.
Everyone just waking up in this video
I just woke up and this was playing. Now I better just know how to program a new LLM or I’ll be really upset.
This is super fascinating
Wonderful, elegant explanations! This is the way to present the basics of a hugely scalable system!
Right lol you and at and I yyyg
@@dannyfrost2621 an me
@@HassanAhmed-rf9xrk,
W😅 w😮😮 1:11 1:11 1:12 1:12 was
@@HassanAhmed-rf9xra
What should be confusing is your vision or at least your conclusion! but the global context is great
RUclips really just teach me Neutral Networks while I’m asleep.
Just woke up after it ended. I remember waking up for a few seconds thinking that it was interesting, hitting repeat and fell asleep after a few seconds again as it was still in the middle of the night.
Now I'm wide awake and hit replay again to truely watch it haha 😂
Why have so few actually chosen to watch this video, I woke up at 4am to it playing
This is such a great tutorial! Thank you for making it. I will share the video with students interested in neural net and deep neural networks.
Uu
This is a lot of videos smashed into one. Honestly, excellent work.
Where would he place GPT4 on his generality performance graph? Must be a step change
If you can us any shape coulad you use a Golden Torus mandala or hipnotic eye. I would try a this shape but it would Spiraled Like a slinky within a slinky?
That's strange, I just woke up exactly 7:30 am to this video playing. And I went to the comment section and turns out I wasn't the only one. I don't remember watching science related video neither.
Marvelous work! If this captivates you, there's a book with similar themes you’ll want to explore. "From Bytes to Consciousness: A Comprehensive Guide to Artificial Intelligence" by Stuart Mills
Nicely explained. One thing near the start is that sigmoid only goes from 0 to 1 (It's tanh (x) that goes -1 to +1)
They r related ,, nice point to mention
Very clear.....thank you
One of those videos you get hooked to when backed
Simple and intuitive explanations. Thanks!
Exactly like yours.
@@wolfisraging Thanks bud!
D@@MachineLearningwithPhil
r@@MachineLearningwithPhil ssss
r@@MachineLearningwithPhil ssss
2:27:45 So your neural network has a neural network lol... I was thinking along similar lines right before you said this... "It could translate to the "rough doesn't fool a human" and then be "translated" again by a network that understands the specific language quirks better and has been trained on natural speech... But the straight intermediate (Latin? Lol) sounds better... I don't guess it even has to be a real existing language even, so long as it is set up to minimize "lost in translation" errors.
You've got one of the clearer presentation methods for me... This stuff is really making sense now finally lol
Thank you for the video :)
This is great.learning neural networks while I drive to work. The internet is beautiful 😍.
Brandon; great video! where can we find more visual representations of adding curves? @40:00 you begin to combine curves. how and where does one learn more?
Fourier series comes to mind. Basically, add a bunch of simple but different curves together to get one complicated but continuous curves.
@@user-bb9lx9gu7c Fourier ne fout rien à la fourrière.
@@whannabi😂
Woke up to this and is exactly what i searched for yesterday, but couldn't find it
I am also here on complete purpose. And can’t stop watching.
Thanks!
I woke up to this like everyone else, apparently.
I'm guessing the unusually long runtime increases the likelihood that someone would, as opposed to waking up on some random 20-minute video 🤷🏻
we both woke up to this
@@purpls.We all did
Nah that can’t be, not every 3 hour video has everyone waking up to it
Wishing you goodluck keep it up 🙏
much respect
Fantastic explanation 👌
I just woke up and turned my phone on to this??
Just woke up. Saw this in my recommended section and immediately clicked, because I didn't understand the datascience introductory course I had.
Fell asleep to a Mr Beast video, greeted by a full in-depth understanding of how neural networks work when I woke up
Please make a video on computer networking and competitive programming
so glad im not the only one who woke up to this vid in the middle of the night lol
Woke up 2 hours 44 minutes in , i am CONFUSHON
7:40. The bottom right neuron is supposed to be inverted. 2 black on top and 2 white on the bottom. The negative weights should actually be positive weights.
Exactly
Exactly
Exactly
Exactly
Would the vectors need assigned values/weights in that case? Isn’t this just an example?
i clicked on this because i was curious just to discover that i watched the entire thing sometime when i was sleeping
I was not watching this when I feel asleep.... but I did wake up to it
Thanks for sharing
Watching this before the TensorFlow tutorial.....
46:28 in b those have a corelation too
BEST CHANNEL, THANKS FOR FREE STUFF.
So if they are saying the jet stream will straighten that intern will reduce the strength of low pressure systems.
Also if the jet stream drops lower that will then make more areas cooler.
P.s the golf stream is not the same as the conveyor belt
3:33 How come the narrater's cough makes the result go from negative 0.075 to positive 0.075?
abs(0.075)
shutup
@@duoko98 Are you saying that because your mother has three? Get an education
@@ansowarrower5038 Lol cough joke just wasn't funny to me idk...
Guys how about a Tensorflow tutorial in depth! Please?!!
Check out the "TensorFlow Basics to Mastery" Coursera Course - www.deeplearning.ai/tensorflow-from-basics-to-mastery/
I am currently doing Course 1. Doing courses separately is free. The specialization is paid.
@@1ycx I think you misunderstood something. You have to pay for a certificate. You can do courses separately but these courses come under specialization only. You get certificate after every course but you have to pay for that. If you don't want certificate then only these courses are free.
Check recent uploads
@@shwetagoyal9801 I think that Koga Master is talking about the fact that you are able to learn TensorFlow on Coursera for free rather than getting a certificate for it.
@@1ycx ini 0 88
i was thinking who in the world would watch something like this 3.5 million times
So super quick, the brain is measuring space time and aligning to it and the more it has to cross reference and integrate, the more data to pull on and weigh the differences, the more aligned the brain is to the current. So the images and thoughts are measurements so not entirely false. If it’s a measurement then we align more to be more accurate. We aren’t really finding “unreal” that’s just a word that means it isn’t current not not existent. Like a chair in a tree or a child before it’s born. Potentials waiting to be exposed by measurements and then reflected through the portions we can bring it out. I know it sounds crazy but the images you see in the head aren’t real but can be pushed that direction through the measurements and reflecting them through our actions and such.
Fantastic, thank you.
1:20:40 Since this course is about learning algorithms this is important to classification. Vector:
noun
Mathematics.
1: a quantity possessing both magnitude and direction, represented by an arrow the direction of which indicates the direction of the quantity and the length of which is proportional to the magnitude.
2: such a quantity with the additional requirement that such quantities obey the parallelogram law of addition.
3: such a quantity with the additional requirement that such quantities are to transform in a particular way under changes of the coordinate system.
Biology.
1: an insect or other organism that transmits a pathogenic fungus, virus, bacterium, etc.
2: any agent that acts as a carrier or transporter, as a virus or plasmid that conveys a genetically engineered DNA segment into a host cell.
Computers.
1: an array of data ordered such that individual items can be located with a single index or subscript.
verb (used with object)
Aeronautics.
1: to guide (an aircraft) in flight by issuing appropriate headings.
Aerospace.
1: to change the direction of (the thrust of a jet or rocket engine) in order to steer the craft.
...
I am tempted to say Physics would include a unit
Physics
1: a quantity possessing magnitude, direction and unit.
I🏵️🌪️🏜️🌿🤹🧚🦹🧚🧚🧜🙍🛌🙍🏜️😚🏜️🌿🏜️🛌🤼🙍🏄🤦🏇😈🛌🙍🤦🙍🚣🏄🤦🤦🤦🏄🙍🙍🤦🌪️🌿🌻🍃🍃🐯🦕🐯🏇🏇🤺🤺🏋️🤵😉🤭
77 por
Thanks but most of us finished high school too 😅
Oo😊
If only life was like that one episode of Dexter. I’d be a quantum physicist by now
Seems like neural net teaching comes in two forms - math or python libraries. It'd be cool to see an intermediate form - some code that implements the math but isn't a library.
ruclips.net/video/aircAruvnKk/видео.html
And one more from dayako
@@seriouscoder1727 Super good video. 3blue1brown is really good at everything mathematics. And a good teacher.
Ong so true not everyone likes python
All of the libraries mentioned are open source, meaning that you can go read the source code. Honestly, though, understanding the content is going to be much more difficult with the source code because it would take away several levels of abstraction that are implementing highly technical details.
Doing this, for example, with Pandas or numpy (or worse, with straight python) would take quite a long time and be useful to almost no one, negating the investment into a video.
@@trevortrevose9124Packages are also available in R, C#, and several other languages, friend. Python is for sure the easiest and most popular general purpose programming language to use, though.
Amazing video, thanks!
My auto play was turned off when I went to sleep and yet, somehow, I woke up to this playing. Not sure if I turned it on in my sleep or something but then again it seems I’m not alone in the endeavor
I woke up and this was on my phone lol
I just woke up.. went down watching vertasium
Haha.. I’ve been falling to sleep to this video for a couple weeks now. Trying to get through it full consciousness. Heavy stuff.. haha
41:48 imagine it in n-dimension
It can do amazing stuff in a matter of second.
Can you explain neural net in time series please
These lessons touch my heart and for the first time i can imagine abit what is going on in that black box
I have a project where I'm taking some low resolution monochrome sprites and I'm attempting to reduce the shape of these sprites in a higher resolution. I tried to play with some scale filters to aid me in that process. It seems a lot like trying to make a vector font from an 8x8 character bitmap.
@3.56, the activation function shown is tanh (-1 to 1) not a sigmoid/logistic (0 to 1).
but the activation function can be anything right!!! It can be anything related to what the input is all about like: ReLU, Sigmoid, tanh, etc...
Essentially, the sigmoid function is equivalent to a tanh function just multiplied by some factor (as well as the argument)
I'm only a minute or two in but this is amazing... I'm understanding it so well that I'm considering learning to code
⁰4
@@vinobrien5738t😅aaà
For busy students: play at 1.25 or 1.5 speed
For busier students: ctrl + w
Woke up to watch 😮
This was on with no way to stop it for several minutes. I think Murphy was trolling me today
I fell asleep with my PC on When I woke up I saw this. Interesting. I did take differential calculus in college and programming for other reasons. Now I have an understanding of neural networks.
Yo!Same thing happened to me,awesome
@@TokyobuckettsLive yes not similar but i was charging my portable charger a left the vids on so it could charge. Came back later an saw this thought it was interesting an now im here hehe.
Is that what you call "deep learning"? ;-)
3 am be hitting when thes videos show up on my feed
First 30s wrong example but good video in general. Yes, you can simply detect what pixels are dark, what are lighter, there is function for that. Also 4 pixels is bad example because you could create like 8 if statements define tolerance (same tolerance as in Photoshop which is distance between colors) and it would outperform every model.
I, too, just woke up to this video - after a nap while on holiday. Lol
This already looks like an interesting course , thanks FCC !!
Yes something to help me with my deep learning course thanks 🙏
1 am and he finally hits me with the human level intelligence section
@7:40 at the last bottom-right neuron on the third layer; shouldn’t the connected weights be positive (white) to get the desired output of the horizontal pixels?
I woke up with the video already finished darn it
A simple rule definitely could solve the first problem. I'm not sure that's a very compelling argument to use a neural network in that case.
shut up man
@@G83X Hey, in retrospect I see that he's just using a super simple example as a point to teach. Which is totally fair.
What are the pre-requisites for watching videos on Machine learning?
a brain
@@yumbemelvin2981 🤣🤣🤣
Abit linear algebra(matrix) abit statistic high schol in enough
Yt and internet 3g
So in slept with the video 1 trillion lions vs the sun and now I woke up 1:38:12 into this video