AMMI 2022 Course "Geometric Deep Learning" - Lecture 1 (Introduction) - Michael Bronstein

Video recording of the course "Geometric Deep Learning" taught in the African Master in Machine Intelligence in July 2022 by Michael Bronstein (Oxford), Joan Bruna (NYU), Taco Cohen (Qualcomm), and Petar Veličković (DeepMind)
Lecture 1: Symmetry through the centuries • First neural networks and the “Perceptron affair” • The curse of dimensionality • First geometric architectures: neocognitron and CNNs • Chemical precursors of GNNs • Geometric deep learning blueprint • The "5G" of Geometric deep learning • Course outline
Slides: www.dropbox.co...
Additional materials: www.geometricdeeplearning.com

[@bithigh8301]

Mr. Bronstein, thank you very much for making it public available. Graph comunity should be greateful for all the effort you have done so far. Keep pushing! Thank you.

[@MichaelBronsteinGDL]

Thanks!

[@adityamwagh]

I like how you showed everyone’s faces alongside their contributions. So wholesome!

[@ujjawalpanchal]

17:18. Symmetry is very close to ML. Because ML is fundamentally based on the idea of pattern matching.
Symmetry is derived even in a dictionary as, ``similarity or exact correspondence between different things". (From Google).
This idea of similar/exact correspondences between different things is verily pattern matching.

[@maxusarron5827]

Do you have an estimate of when the Geometric Deep Learning book will be published? Also, can I know how much it'll be extended compared to the protobook (150 pages-ish)?

[@MichaelBronsteinGDL]

In 2023. We expect it to be at least double the size and more pedagogic and self-contained

[@maxusarron5827]

@@MichaelBronsteinGDL That's great to know. I'll be waiting patiently for it and I'm sure the rest of the GDL enthusiasts will be as well. Thank you!

[@SajithMR]

Thanks for the great lecture.

[@Computing_Brain]

01:01:28 GDL Blueprint

[@mengni4426]

Thanks for sharing. Open a window for me!

[@syedsajjadalishah2011]

find the whole playlist impressive, sir kindly share your official github link

[@samabenojar49]

I want to learn this type of stuff. Where do I start?

[@MichaelBronsteinGDL]

Our proto-book on which the course is based in a good starting point: arxiv.org/abs/2104.13478

[@mujtabaalam5907]

@@MichaelBronsteinGDL What sort of mathematical background is required? Recommended? It seems to use quite a bit of signal/information theory, group theory, and differential geometry

[@sankafoodey]

Group Theory, Graph Theory and Differential Geometry are the real deal in this course.

[@mujtabaalam5907]

5:31 how do we know he was likely aware of Khayyam's work? They lived hundreds of years apart and didn't speak the same language

[@MichaelBronsteinGDL]

This is not certain. He used a similar construction (quadrilateral)

[@mujtabaalam5907]

@@MichaelBronsteinGDL Thank you.

[@alivecoding4995]

I think the book really needs some more refinement. I found it unnecessarily "difficult" to read, as it does not explain the essential parts in detail (mathematical results mainly), while explaining other parts too lengthy. Especially the Figures are often not that helpful. Also, the type font for the group theory isn't well chosen and hard to read. I really cannot recommend the book in its current state.

[@MichaelBronsteinGDL]

We are working on it, thanks for your feedback

[@alivecoding4995]

@@MichaelBronsteinGDL Highly appreciated! Your overall efforts are great. I am thankful.