Compressed Sensing: Overview

Рет қаралды 58,824

3 жыл бұрын

This video introduces compressed sensing, which is an exciting new branch of applied mathematics, making it possible to reconstruct full images from a random subset of the pixels. There is a ton of beautiful math behind this concept, touching on high-dimensional geometry, robust statistics, and optimization.
Book Website: databookuw.com
Book PDF: databookuw.com/databook.pdf
These lectures follow Chapter 3 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

Пікірлер: 67

@elromulous 3 жыл бұрын

It's rare to find a lecturer who explains complex topics so lucidly! Thank you!

@zieglerai 3 жыл бұрын

I have been working with compressed sensing in my bachelor thesis and a few months ago I couldn't find a collection of videos explaining CS like you are doing here very easily and interactively. Thank you for taking the time to make and upload this great content!

@MissyMcFruit 3 жыл бұрын

These videos are so helpful as I am writing my master's thesis in Compressed Sensing! So many books and articles are so technical, that it's hard to see the woods for the trees. Love your teaching style!

@filipee77 3 жыл бұрын

Amazing. I use CS in my PhD here from Brazil. Great explanation and work.

@Eigensteve 3 жыл бұрын

Thanks!

@somethingnew7538 3 жыл бұрын

I really love the way you teach, And The famous line "so on and so forth".♥️

@Kracheta 2 жыл бұрын

OMG this is my new favourite channel! You explain so well, thank you!

@mohammedsaad7122 Жыл бұрын

first of all thank you so much for the in-depth explanation. This is literally whats gettting me through Compressed sensing Class. Secondly, its mind blowing to see how easy mirrored writing is for you.

@yasir9909 3 жыл бұрын

You have presented a very good, concise, and easy-to-understand description of the subject matter for an average student...

@Eigensteve 3 жыл бұрын

Thanks!

@medad5413 2 жыл бұрын

Thank you for making the book and code available.

@EduardoGarcia-tv2fc 4 жыл бұрын

What a great book and series!

@zonglehuang4147 Жыл бұрын

Video with high quality! Really helps a lot.

@franciscojavierramirezaren4722 3 жыл бұрын

Thanx for sharing, great book! Greetings from México!🙂

@yanhairen7293 3 жыл бұрын

This lecture is so good. Enjoy it so much

@slevon28 3 жыл бұрын

Great job. Your book will definitely be a reference in my PhD-Thesis! Thank you very much.

@yaakoubberrgio5271 3 жыл бұрын

Hello can you find me matlab code of compressed sensing 1d and 2d Thanks

@Veptis 2 жыл бұрын

I wonder if I any course at my university touched on this topic. If not I might watch all the material on your channel. It's really interesting as I remember the idea from an old forum post for the use of thermal imaging. Which is quite low resolution and really noisy. Also impressive to see you writing everything mirrored and making it look effortless

@fzigunov 3 жыл бұрын

Can't wait to see the math!!

@MaksymCzech 3 жыл бұрын

Great stuff, just like always!

@pablogiuliani4618 2 жыл бұрын

This is fantastic, thank you so much Steve and Nathan from the massive amount of work that went into writing the book and making these videos. BTW, I was wondering if you learned to write in an inverted (flipped) fashion just for the sake of these videos, but then I thought most likely you wrote normally and then flipped the video itself. The position of your ring says that is most likely the correct answer ;).

@TheKhakPakflyest 3 жыл бұрын

Fantastic as usual.

@greje656 3 жыл бұрын

Really interesting. You are a fantastic presenter btw :)

@zhihuachen3613 3 жыл бұрын

really useful, thank you professor

@craigcollings5568 2 жыл бұрын

Databook downloaded. Also, Sparse identification of nonlinear dynamical systems

@mariahanveiga2478 3 жыл бұрын

your videos are so great!

@rickharold7884 3 жыл бұрын

Can’t wait ! Thx

@odlz3636 3 жыл бұрын

Many thanks !

@manuelpena3988 3 жыл бұрын

This teaser really has me waiting for the next video. It never happened to me with a movie xD

@baranidharanv8648 10 ай бұрын

Its really useful

@neb5615 3 жыл бұрын

Thank you

@hrtlsbstrd 3 жыл бұрын

Love your communication style, a lot of professors could do well to learn from you.

@duffahtolla 3 жыл бұрын

I wonder if this could give you a preview of long exposure digital photographs used in astronomy.

@anilkumar-kv4we 3 жыл бұрын

Thank you sir

@osten222312 3 жыл бұрын

thanks!

@alfonsollanes888 Жыл бұрын

Magnificent lectures about image compression. I have a question about extracting distance between and within images from a flat photo image. Do we have the projective geometry math available for this process or is flat photography need to depend on technology like LINAR?

@Blakut 3 жыл бұрын

He circled s using the l-inf norm

@existenence3305 3 жыл бұрын

Turns out, Steve is an AGI that came from the future to teach us about how to build him.

@taimurzaman7322 Жыл бұрын

Damn it, you gona giving me my Ph.D. so easily :) see how his books are available for free.

@zrmsraggot 2 жыл бұрын

Hello, could it be possible to do it within a movie ? Having a representation of the climate with only a few measurements over a continent and then being able to predict how it will evolve would be great ! Also I wonder if randomness is optimal for Compressed Sensing, is there a way to tell if some measurements are better than others ? Thanks to anyone who read this and takes time to answer

@pepaxxxsvinka3379 3 жыл бұрын

Nice video! I am trying to understand this method. Can I resize the image using CS?

@mamatoshgupta 2 жыл бұрын

Re-frame your problem as follows: The image that you have is a sparse sampling of a higher resolution image which you want (resized image). Then apply the CS reconstruction theory with the appropriate compression matrix.

@INDIANHeroes 3 жыл бұрын

Does compressed sensing uses matrix multiplication anywhere in its implementation?

@Eigensteve 3 жыл бұрын

A lot of the algorithms involve iterative solvers, each step of which often involve some sort of matrix inversion. Can get pretty expensive.

@kansasmypie6466 2 жыл бұрын

I can’t believe all of this is for free. That’s too good to be true lol

@EEDKonduruLakshmiBhanuPrakashR 3 жыл бұрын

hi sir, as specified in video X=basis function(Fourier)*S(sparse) that sparse matrix contains already the less number of non zero pixels and maximum of zeros. how we will get that sparse matrix vector of image the multiplication of basis and sparse is equal to X. so that we can write it as Y = MEASUREMENT MATRIX)*(BASIS_FUNCTION)*(SPARSE IMAGE) i am confusing how to get that sparse image and how to design measurement matrix professor can you please clarify my doubts. thank you professor

@prashantsharmastunning 3 жыл бұрын

if we use actual row resolution image in place of randomly sparse matrix,. will we be able to high res it ?

@yaakoubberrgio5271 3 жыл бұрын

Hello Can you find me matlab code of compressed sensing Thanks

@er.dharnabakotra8849 3 жыл бұрын

Respected sir, May I have MATLAB code for practical purpose for checking and making image sparse?

@kopapa 2 жыл бұрын

Thanks for this compact overview. Btw: That “Infinitely many ass*s” made me giggle.

@INDIANHeroes 3 жыл бұрын

Can anyone tell how to verify the results obtained from compressed sensing are how much accurate ?

@sj.j5169 3 жыл бұрын

Is it necessary that y be a [p , 1] vector or one could use y as a [p, l] matrix? If I had a two dimensional matrix as y, should I re-arrange it as a long vector or use it directly?

@charismaticaazim Жыл бұрын

From what i understand, its a matter of convenience in computation, due to which its converted to vector (1d). Is this convenience necessary ? For now, my guess would be yes, to ensure compatibility.

@yaakoubberrgio5271 3 жыл бұрын

Hello steve Can you find me matlab code of compressed sensing Thanks

@jajula2565 3 жыл бұрын

@yaakoub Berrgio can u share ur mail contact to reach u jsrece87@gmail.com

@convex9345 3 жыл бұрын

Unable to download the pdf and the book is not available in India , what is the solution?

@chinhung6928 3 жыл бұрын

Why not L0-norm of S ?

@dibyalekhanayak2130 2 жыл бұрын

Hi sir can you pls give some description about omp based reconstruction

@dibyalekhanayak2130 2 жыл бұрын

Can any video of your is covering basic of deep learning if possible pls give the link

@nickey0207 3 жыл бұрын

It looks like your lower body is not in R(A) in the video. We might recover it tho.

@iaggocapitanio7909 2 жыл бұрын

teaching functional analysis as simple like that

@elgracko 2 жыл бұрын

Engines of our ingenuity brought me here, 🤙

@mohamedaboelenein7727 7 ай бұрын

So basically, magic!

@alexanderskusnov5119 3 жыл бұрын

The same effect in the piece of hologram.

@jamescarson668 Жыл бұрын

Love this series. @SteveBrunton is awesome! Just want to throw out to be helpful, "Beg the question" should be "raises the question." The former is a way of pointing out a fallacy: en.wikipedia.org/wiki/Begging_the_question. I used to make the same mistake and was happy when someone corrected me.