SVD: Image Compression [Matlab]

This video describes how to use the singular value decomposition (SVD) for image compression in Matlab.

These lectures follow Chapter 1 from: "Data-Driven Science and Engineering: Machine Learning, Dynamical Systems, and Control" by Brunton and Kutz

This video was produced at the University of Washington

[AlessandroBottoni]

Absolutely amazing! This video series should be considered a "national treasure" and kept safe at the Congress Library. Congratulation Prof. Brunton and many thanks for you valuable work.

[matheusparanahiba]

I'm learning Recommender Systems, so I came across the concept of SVD and searched for a video to better understand it. I couldn't be happier to find this amazing series, which has been most helpful and very enriching. Thank you so much, prof., regards from Brazil :)

[rudypieplenbosch]

Wow not only he provides us with brilliant lessons, he even shares his book in PDF form

[AnimationsJungle]

Sir you are simply the best. I have learned a lot from your lectures. You are simply becoming my role model in my phd lots of love from kashmir

[SandeepSingh-yxsi]

Amazing Steve. You have really helped me in understanding SVD application in Data Science. Thanks a lot. I really hope if you could have made your videos downloadable

[franciscogaray]

Simply spectacular your way of teaching....you are a great teacher, greetings from Perú, South America

[raviprakash]

Excellent lectures. I was having trouble understanding SVD. This lecture helped me a lot.
Thank you very much for uploading.

[hongwang]

Dr. Brunton, thanks very much for your excellent lectures!

[BoZhaoengineering]

thank you for your video and PDF books. Data science is now everywhere. Your video and the book alongside are my ongoing resources to visit, when I need a certain mathematic technique. I work for wind power section as a structural/mechanic engineer (of course a math lover). For the topics of Load Simulation, various of vibration and aerodynamics such as turbulence are the ones I am working on daily. Cheers,

[rasher]

Excellent lecture series!!! This is really inspiring and probably the best lecture series ever. Totally transforming the way we look at SVD and its applications in real life. Thank you! for your efforts and passion to create such lovely teaching materials, including the other lecture series on machine learning, control systems, and data-driven dynamical models.

[cxxocm]

I was trying to understand PCA and Googled this amazing series. Thanks, Dr. Brunton. Not only the contents and explanations are stunning, but also the technologies used in the lectures were fabulous. The only complaint is that sometimes I couldn't focus because I'm thinking that how could Dr. Brunton write reversely? What fancy technology he was using? :)

[jacobanderson]

thanks for positing these. Definitely buying your book!

[HavaNrus]

11:22 What would also be great is too add here Frobenius norm error graph, and show it's decreasing.
Also I have a question about hidden watermarks you talked about: If I add a big enough watermark even to parts related to the last eigenvalue, wouldn't it change the whole SVD basis?
Btw, thank you, you're lectures are God's blessing on mankind. 👍

[andrezabona]

Professor Steve, why people use FFT or Wavalets instead of SVD? For which application is SVD approach better then other two?

[SoumilSahu]

Just to make sure I've understood this correctly, since you're performing the SVD for a single image, you're essentially seeing how well the "pixel columns" of the same image are correlated to each other, correct?

P.S. the idea of digital watermarking seems so simple yet so cool, this is amazing stuff!

[kasturibarkataki]

Really really grateful to you for helping me learn this!

[syoudipta]

With just 5 modes, you can get a "Ruff" estimate!

[Martin-iwll]

Yes, good to know you are a fan of Terry Prachett as well

[noahbarrow]

Steve (and co.) I am a huge fan. You've deepened my appreciation of linear algebra, data science, fluid mechanics and matlab itself. Thank you! I recently purchased your book and I haul it around with me to school like it's one of the dead sea scrolls. I'm trying to better understand this idea of cumulative energy...can it be thought of as the 'effective power' of the rank of our sigma matrix? In this video with the image of your dog it appears that we approach the energy of rank 1 as we include more information, right? Am I understanding these nuances correctly? Thanks again for all of these videos. The clarity, passion and enthusiasm you have for these subjects is inspiring!

[qamarkilani]

I cannot wait for the next lecture .. Very informative .