Learning to See [Part 11: Haystacks on Haystacks]
Вставка
- Опубліковано 11 чер 2024
- In this series, we'll explore the complex landscape of machine learning and artificial intelligence through one example from the field of computer vision: using a decision tree to count the number of fingers in an image. It's gonna be crazy.
Supporting Code: github.com/stephencwelch/Lear...
welchlabs.com
@welchlabs - Наука та технологія
This series is probably the best stuff you can find in youtube right now. Thanks and Congratulations!
Thanks for watching!
I really like to see the entire travel from the problem to the solution! This is perfect to me!
Those videos are so good and emotionally engaging that every time one ends I cry a little. Great job!
Ha, thanks for the encouragement. I've been working on a new script. Your videos are great!
It's okay, we can wait for few more years for hackerdashery's video.
Woah - just saw that you commented! Awesome! Best of luck with the new script!
@@RazinShaikh still waiting :(
@@unflexian Still waiting :(
Literally this morning I decided to go and learn about one of the Millennium Problems, and decided to choose the P vs. NP problem. And then this video comes and mentions it. Coincidences are weird.
Cool!
Hope that it does not creep you out but t's no coincidence
This series, and really, all your other work, is just phenomenal. Thank you! And please keep it up.
What's weird is, I already pretty much know the answers to all these problems. I've never appreciated how ridiculously awesome that algorithm is, how impossible the scale of the problems it brings in our computational reach, how recent those mathematical innovations are.
That's exactly how I felt researching the series! Thanks for watching!
Okay, every time I see one of your videos show up I am both simultaneously extremely excited and extremely scared. Excited because I get an answer to questions I have been thinking of for weeks, and scared because you throw a tonne more at me. Being interested and being left in the dark at the same time is god awful.
agreed
Welcome to academia.
That is so fascinating! Can't wait for next episode.
Good job!
My sadness grows exponentially with the forward movement of the seek bar. Because I have to wait 2 more weeks for the next one. I wish these videos were never ending. Hats off to you! I am truly inspired by your videos and currently making contents on machine learning in my native language Bengali, people are loving it. Thanks again, keep making these awesome videos. :)
Awesome, thanks for watching!
Made my day with this playlist...
Thank you!
This is a great series of videos. I can't wait for the next one! It's clear that a huge amount of time and effort goes into each one.
Most accessible and interesting Intro to Machine Learning on UA-cam.
I like the music 💘
Thanks!
I know little to nothing about programming, but your videos are just so interesting. I always wondered how those decision trees actually worked
I need the next episode NOW and btw love your work
NP = no problem
non-deterministic polynomial in case someone got deceived and actually want to know what it means
Actually it's "Non-deterministic, Polynomial time", meaning NP is the class of problems solvable in polynomial time by a non-deterministic Turing machine.
It is great to see you use such elegant example and decision tree to illustrate key ideas in supervised learning!
I think your videos can replace a major part of the undergraduate level machine learning courses in almost all universities.
I look forward to more of your videos. It serves a great education purpose for me, even if I took 3 ML-related courses at CMU, and just got a PhD :)
Wow, thanks for watching!
I think for the next series I will wait for it to finish these cliffhangers start to bother me to much
exactly what i was thinking
Thanks for the wonderful videos you have been producing. Thanks a lot man. Your videos are short, straight to the point and above all it is not boring to watch. Please can you start similar course on the python programming language.
Thanks for watching!
still one of my favorite series!
This channel is way underrated
These series are done so professionally. Well done on the great content. Keep up the awesome and amazing work. :-)
Thank you! Was waiting for this video.
Great quality as always!
yay new episode!!!!!!
Thank you! Can't wait for the next video
I literally *JUST* realized this channel is actually about programming
it is?
Brilliant, as usual...
Wow, it will actually take more than a month for 6 pixels :o. Great video as always !
you are incredible!
A big fan from Mesopotamia ❤
God, you're videos are really good and i dont know how to handle it
amazing series! ill start my CS degree in September!
dang. waiting for next saturday already
omg...don't make us wait 8000 yr! :(
I can not handle this cliff hangers
Oh I love you guys.
I realy like your work and hope to give you a littel tip by buying your Imaginary Numbers are Real book. love to see the next video
Appreciated! Thanks for watching!
cant wait!!!
0 Dislikes and lots of nice comments. And you totally deserved it!
It's been awhile since university, but what I remember is that the traveling salesman problem isn't "what is the shortest route," but rather "given a specific route, can you determine if it is the shortest?" Still NP-complete, but I think a little easier than what you presented it as.
7:35 wonderful
What's the song in the beginning? 😍 love it!
Well, if you tested a set of pixels then you can be certain that checking almost the same set of pixels will yield an almost identical result. So we don't have to test all pixel combinations, but enough combinations that are different to some specified degree.
Next topic: measuring cuteness.
Good for you "Max the Potato"!
nice catch!
Thanks!
Cliff hanger!
The answer: some pieces of hay are sharp enough to be considered a needle! (In other words, the same rule applies to the "rules" that you test as do most of the pixels in the screen--many can and should be ignored)
Using a magnet!
What kind of infrared camera are you using? I can't find a cheap one that is small, like yours.
He is using Leap Motion... He mentioned it in the first episode.
NP != exponential as you show in you video. NP is nondeterministic POLYNOMIAL. So it solvable in polynomial time by nondeterministic Turing Machines, there are problems not solvable in polynomial time even for nondeterministic TM they are exponential. So we definitely know that P = NP (and hence P == NP, and majority of mathematician believe P < NP), i.e. for every NP problem exist polynomial time algorithm for deterministic TM. So as you are right to bring attention to NP problems it seem mentioning them is irrelevant to what you show and your passage leads to wrong assumption about NP == exp.
P.S. I use less, greater and other combination for relation of being subset (strict or not).
P.P.S. I love your incredible good series so much!
Given the quantity of rules, why can't we simply have rules based on "if this rule, then test this rule"? For example, we could find the best 1-pixel rule, and search for an alternate 1-pixel rule if that pixel isn't active, or search for a 2-pixel rule that uses the pixel from earlier if said pixel is active. This way, it's been turned into a hunt for 1-pixel rules, rather than anything more complex.
such a cliffhanger :-p
:)
LIKE!
Intro song/nusic:
www.premiumbeat.com/royalty_free_music/songs/what-s-the-matter
where is the number 7867 coming from? is it from the image ?
You and your cliff hangers...
Awwwwww.........
(About the time)
The series is interesting, yes... but i think it seems like never getting to the point. Is the series NP?
Luis Miguel Gallego Gómez my god it's turing's uncertainty principle applied to a video
exactly my thought. the videos are very nice, but it just drags on forever unnecessarily....
Well any other explanation is just going straight to the end so you don't understand anything.
I really like the pace and completeness of the explanation.
If he made it shorter, the series would have to sacrifice in pace or completeness.
I think he is trying to explain everything so that a viewer with no knowledge on the subject can understand what is going on
Are you studying or working ? if studying , which college ?
I hate that I have to wait so I can learn, this never happens!
Anast Gramm exactly
What IDE do you use for Python?
Jupyter..
check here : jupyter.org/
Hold up. Why is Tetris NP? What is "the Tetris problem"?
Are you sick? Voice sounds a bit different, but if you're not glad you're okay. If you are hope you feel better soon.
Jerome H. *Friedman* 7:57
That makes me think...if a self-aware AI wants to see, it's task is not to try to spot the one thing..then another thing...but generally it wants to know What information can it obtain if it was able to see.
...am I doing the whole "generalization" thing right?
Use a magnet!? Right?
The answer is, a very large magnet...
*Deterministic
0:20 roughly 1.6 seconds short of an hour. nice
yeah, that really worked out nicely!
What is this business about Mr Hackerdashery at 9:48?
Just that I would like him to make more videos!
You use a magnet
Hahaha good call.
You are such a tease. But, change nothing.
first
just burn the hay and use a magnet......
I've found best cbd products on WeedBorn.
I guess he tried to make a series that matches every viewer.
I assumes that People don't have any Idea of programming or basic math AND
no time to watch more than one video a week AND are rather slow thinkers.
You have to be realistic about your target audience.
You will never get anywhere without letting some of these groups down.
And by the way, you've already lost everyone who doesn't already knew a lot of this stuff.
So you might as well just go a little faster. And don't repeat everything forever.
If someone needs repetition he can just watch the videos again!
alexander kerbers
I personally love the pacing of these videos. It feels calming and interesting af.