Brad Osgood's lectures are phenomenal. He was trained as a mathematician, so his explanations are crystal clear in terms of definitions and where physicists sweep mathematical formalisms under the rug. After watching his entire lecture for this course, and spending a lot of time reading the course reader, I learned a lot about how to decrypt the kind of mathematical nonsense that physicists tend to say (as well as learning about the real foundations, c.f. tempered distributions and others).
That being said, a lot of these lectures can be summarized in a few sentences if you already have a strong foundation in linear algebra. For example, the fact that complex exponentials form an orthonormal basis for periodic functions is the content of the first handful of lectures, and deriving the Fourier transform is only slightly messier to explain in terms of linear algebra (this is not because of the linear algebra but because Fourier transforms are inherently a little messy).
The Fourier transform is possibly my favorite mathematical function, and the closest I've ever been to "Whoa, math is beautiful". Ironically enough, I reached this conclusion in a signal analysis class, not a math class.
I'm interested in understanding more about the use of the transform in discrete signal processing. I wish I had the time to try and absorb 30 50min lectures, but I probably don't.
Can anyone recommend a more limited course or tutorial focussing on that, or should I try and extract particular lectures from this course?
Yes, I remember that well. Seeing the animated graphics linked within that article really was an 'ah ha!' moment for me; I'd definitely recommend it as an introductory intuitive motivation.
Thanks for the link. His website is terrific. He offers some perspectives for gaining intuition about complex numbers that I'd never heard before, even after all these years. This should be required reading for HS math students.
These are excellent lectures! I saw Chap 4, and it was very satisfying to watch the introduction to convergence, L2 norm, inner product etc. Technically, Fourier transform requires only the L1 norm (the 'Manhattan distance') to be finite, but L2 intersection L1 is a superb way to motivate distributions. These lectures must be made required viewing for all students of engineering and mathematical sciences.
Also from Stanford, from longer ago... The Fourier Transform and its Applications by Bracewell. McGraw-Hill. Extremely well written with great examples and some fascinating problems. I highly recommend it. Discrete transforms are a bit of an afterthought in a final chapter, but enough to dig in and see how the FFT works.
In the same sequence, Stephen Boyd's course Introduction to Linear Dynamical Systems (http://see.stanford.edu/see/courseinfo.aspx?coll=17005383-19...) is also excellent. The title sounds pretty specific but it is really about a wide range of applied linear algebra. Boyd's lecture style is a little idiosyncratic but he's great about constantly emphasizing how to gain an intuitive qualitative sense of what everything means.
He also has a couple of courses on convex optimization that cover the theory behind a lot of machine learning.
22 comments
[ 40.5 ms ] story [ 231 ms ] threadThat being said, a lot of these lectures can be summarized in a few sentences if you already have a strong foundation in linear algebra. For example, the fact that complex exponentials form an orthonormal basis for periodic functions is the content of the first handful of lectures, and deriving the Fourier transform is only slightly messier to explain in terms of linear algebra (this is not because of the linear algebra but because Fourier transforms are inherently a little messy).
Can anyone recommend a more limited course or tutorial focussing on that, or should I try and extract particular lectures from this course?
Really great book in any case.
dsp-book.narod.ru/FFTBB/0270_PDF_C16.pdf
http://www.katjaas.nl/home/home.html
This is the one that made me get it in the end.
He also has a couple of courses on convex optimization that cover the theory behind a lot of machine learning.