The Fast Fourier Transform (FFT)
Vložit
- čas přidán 30. 03. 2020
- Here I introduce the Fast Fourier Transform (FFT), which is how we compute the Fourier Transform on a computer. The FFT is one of the most important algorithms of all time.
Book Website: databookuw.com
Book PDF: databookuw.com/databook.pdf
These lectures follow Chapter 2 from:
"Data-Driven Science and Engineering: Machine Learning, Dynamical Systems, and Control" by Brunton and Kutz
Amazon: www.amazon.com/Data-Driven-Sc...
Brunton Website: eigensteve.com
This video was produced at the University of Washington - Věda a technologie
It is so crazy that Gauss discovered a lot of things in mathematics that took people hundreds of years to realize.
Makes you wonder how many people have ideas that could change the world, but choose not to share them because they don't see their full potential (or they assume someone else has already had that idea).
Fun fact 101: when something is not named after Gauss is because somebody rediscovered it later or it would be confusing as everything is already named after him. Probably the latter though.
@@nameismetatoo4591 i think its far more interesting to think how many people could have potentially had great ideas but were just exploited working class people who never had the opportunity to actually form their intellect and study something
@@nameismetatoo4591 Reminds me of the newton-leibniz calculus controversy.
Heavy Gauss Rifle
Very well produced - thank you Steve for this excellent lecture ! FFT is truly what drives the World today... and into the future - with endless applications, in the physical sciences astro, aviation, and medical world.
This is what online lectures should be like. Thank you very much Dr. Brunton for sharing these lectures. I can't emphasise enough how amazingly done these are.
I was just watching this but I kept being distracted and impressed by the fact that you are writing backwards. :O
Ahahaha same here XD
great content in the video but such videos are extremely distracting and make me feel uneasy...I guess any right brained person would find these very distracting.
Same here , you wrote them so naturally without any hesitation
He isn't. The video is mirrored.
He isn't writing it backwards, there is very easy, logical explanation. This has been mirrored, and if you look closely you can see that he has a ring on what would be his right hand, which isn't right, usually rings are on left hand.
This content is amazing, thank you so much for posting this. I knew how to compute a fourier transform of on a defined function but was incredibly confused how computers did it on the sample data they create from analog signals. I had no idea you could do it to discrete data.
No words to express my gratitude for this awesome content
Please Prof. Steve Brunton
kindly we need video lectures on the wavelet transform , DWT , CWT , etc , thanks and best regards
The best lecture series I've seen in CZcams. Thanks a lot for everything.
This format is simply the best.
Easily one of the best instructional videos on CZcams, the clarity in your articulation of the concepts makes the otherwise murky subject so much more approachable. Can't applaud you enough for putting these videos togather. Cheers !
This lecture was like a trailer to the actual one (which I assume comes later in the series). He didn't actually do anything here.
An important point I missed in the video is the Kronecker property for the multivariate case. This enables the use of many 1-dimensional operations instead of one N-dimensional operation. Also called "vec-trick" on tensorproduct elements.
Your Videos are So awesome and wonderfully high quality!
the best series I came across recently
Concepts simplified to the very core. Thank you for the lecture series!
You're welcome!
Amazing Prof Brunton.
In addition to satellite TV, it is cool that the new digital Terrestrial TV broadcasting standard ATSC 3.0, which has just commenced in US also uses OFDM-based modulation and consequently requires FFT blocks on the receiver side and iFFT on the transmitter.
Thank you so much for making this course publicly available professor!
Your approach to teaching Fourier Analysis manages to provide a level of intuition on the subject that makes the equations themselves seem much less daunting.
Also the anecdotes and stories you weave into this course are pretty much the icing on the cake.
I wish there were more black people in Science and mathematics
Thank you so much, I am so excited to learn when I watch your videos!
You are so welcome!
Are we not going to talk about how well this guy writes backwards? 🖊
He writes regularly and the video is mirrored ;)
@@marcnassif2822 Ha. Seems I didn't give that any thought because I _wanted_ it to be true! 😋
@@marcnassif2822 Is he left handed then?
@@akhilezai His handwriting is way too neat for him to be left handed haha, but yes he is left handed.
did my man just casually write on the board backwards for us to see it in the correct orientation? Because that's impressive
Thanks you really rock and you’re a great story teller!!
Thank you so much for these very clear explanations! They are really helpful
Professor, please tell me how can I monetarily support you. The contents you created are beyond brilliant!
I think buying his book might be a very good idea.
Thanks from the lecture!
from Japan
Your welcome from Seattle!
thank u for prompt reply. Be Well !
beautiful video - very well explained
Can't wait to watch the next video...i really love your work
Awesome! Next one should be out on Saturday.
@@Eigensteve Can't wait till Saturday..😄..haven't found any good content on fft algorithm on CZcams..really looking forward to it
Steve ,you are the best .
I wish I could be your student in my uni life 😭 you explained what I need to grasp
Wow! This is an awesome explanation! Down to earth, straight forward, excellent! BTW - you are quickly, and legibly writing backwards like some kind of Leonardo DaVinci !! What the heck! Incredible!
Hey David Cardin, do you like listening to songs by Imagine dragons ?
Fantastic! What system did u use to produce the lecture?
Great video! Thank you!
Do you plan to explain the algorithm and the math behind it? Trying to write this algorithm for a compute shader
Yes, I believe it will come out on Saturday.
really high quality info, thnx.
if I plot the spectral where the X axis is time, do I have to IFFT first? thank you
Wow did this just make me understand scaling the dow Jones day trading ? Very useful information! I wish this guy was my personal teacher!
Dear Prof. Brunton, is FFT mostly used for simple domains problems? (FEM, FVM, Meshless, etc)
It is so crazy that Steve wrote every notes from the back, which means every characters and graphs he is writing should be flipped along y axis by 180 degrees
Obrigado, professor, por nos explicar o porque de usar o FFT (n x long) ao invés do DFT( n x n).
De nada
Brs everywhere.
Um abraço
Are ‘Private Vids’ available under your Membership Plan ?
Sorry about that... that video should be coming out at the very end of this series on FFT, in about a month. Stay tuned!
how does he write backwards so well ???
maybe the video is inverted . He writes normal and then they invert it using software
@@dzemper9410 If he's writing normal then the inversion would be backwards
@@jd87a but the camera sees from behind the board, so inverting again in software will put it correctly
You can search on Lightboard or Lightboard Studio (either of those names) to see more on how this works!
Left handed and the image is inverted.
a) What is this FFT image called in general? (b) What kind of information can you obtain from the FFT image? (c) Is this same as an electron diffraction pattern?
bruh
Sparsity and Compression is a private video... is a part of any membership plan?
Sorry about that... that video should be coming out at the very end of this series on FFT, in about a month. Stay tuned!
Holy shit. Thank you. Thank you so much.
awesome, thanks!
you are too brave keep going!!
I never understand how you do your videos. How the heck do you write in the air, and how you this invisible board trick. Please explain
Where were you all my college life?
I was wondering who invented FFT so I went to wikipedia, letting the video continue to play while I tuned it out to read. When I tuned back into the video, you were just finishing explaining exactly that. Oops 🙃
Nice :)
When we say O(nlog(n)) isn't the log base 2? so in the case where n = 1000, log(n) ~= 10 not 3?
I guess it doesn't matter as much in big O notation because it only conveys a general trend while omitting most of the less significant factors. But yes, Cooley-Tukey FFT is O(n*log_2(n))
awesome video and explanation.... how the heck are you writing backwards??
very good
Thank you so much for explaining complex thing really Easy way!
Can you do this for "Homomorphic Encryption" too??
I'm by no means an expert in encryption, but that would be a fun series.
please help me with this, why for a 10 sec audio, n=4.4x 1000000. what basically 'n' is?
God bless you!
I think in the N Log N, the base is not 10 as mentioned here at 3:30. I think the base should be 2.
I think I agree.
FFT, how about that FHT (Fast Handwriting Transform)??? Can you reveal that algorithm?
probably called mirroring or vertical inversion of video :D
awesome
Amazing explanation! But what I couldn't wrap my head around is how can he write backwards so casually ?!
oh video is inversed on X axis! great move 😉
Ông này viết ngược luôn ghê vch :)) respect!
I just recently read a paper that it's actually faster to just compute the DFT if you're using GPU acceleration, since matrix multiplication is inherently more parallel despite vendors actually providing their own optimized FFT libraries. The performance benefit of DFT is even greater the larger the input compared to the optimized FFT library.
The paper is:
Davuluru, Venkata Salini Priyamvada; Hettiarachchi, Don Lahiru Nirmal; Balster, Eric (2022): Performance Analysis of DFT and FFT Algorithms on Modern GPUs. TechRxiv.
Ok thank you :)
I thought the complexity of FFT was n*log2(n) not with a base of 10?
You can go between log in any bases by multiplying with a constant. So log2(n) = log2(10)*log10(n)
@@Eigensteve but you have no constant in front of the log(n) term in the video. Is the constant just ignored because it is a complexity formula?
@@ceeb830 That's right, we usually drop the constant, since we are just interested in how the trend scales for large n
Gauß was majorly underestimating his own work
So he's left handed, can you figure out how I figured it out?
T-Pain owes his career to FFT
If you can right in reverse, you can explain the Fourier transform.
So this is just an introduction of FFT? Well I was hoping for learning the details and implementation.
Never mind. Found the next video
Isn't it O(n(n+1))?
8 minutes for NOT describing the FFT
idek what ur talking about but nice video!
Karl Friedrich Gauss must have been, no doubt, one of the smartest men who ever walked the earth. Absolute genius.
Did he really write mirrored on glass better than I write normal on paper?
I was watching a video of a kid drinking a bottle of Gatorade through a toilet paper roll straw. How did I end up here?
Hardware is the physics. Software is the math.
That's logN base 2, not base 10. So for n=1000 we'd get logN = 10
@Michael Smith I don't have an idea what you're talking about?!
Gauss was a freak
fft batch
bff2873
Left handed
Don’t watch. He doesn’t explain the FFT.
Please Prof. Steve Brunton
kindly we need video lectures on the wavelet transform , DWT , CWT , etc , thanks and best regards
Coming up soon!