You're reading the public-facing archive of the Category Theory Zulip server.
To join the server you need an invite. Anybody can get an invite by contacting Matteo Capucci at name dot surname at gmail dot com.
For all things related to this archive refer to the same person.
Bruno Gavranović (Oct 04 2022 at 20:43):This is the official topic for the course's first lecture, "Why Category Theory?"
Alec Stashevsky (Oct 10 2022 at 07:37):@Bruno Gavranovic I still haven't received the zoom link for the first lecture. Has that been sent out to our emails or in the google group? I check both places.
Nicolas Dewolf (Oct 10 2022 at 08:19):I also had a question. We received two mails about the first lecture. One mentions it starts at 5pm UK time and the other 4pm UK time. Which one of these is the correct one?
Ieva Cepaite (Oct 10 2022 at 08:20):It’s definitely 4pm UK time :) Bruno confirmed on twitter and I think that’s part of why the update was sent
Ieva Cepaite (Oct 10 2022 at 08:23):As for the link, I imagine time zones might mean it’s better to reach out to other organizers as it’s night time where Bruno is at. I’ll see if I can reach someone else to ask.
@Pim @ @joaogui1 (he/him) @Petar Veličković
Daniel Ordonez (Oct 10 2022 at 08:34):The calendar events are configured in a bad way. I received an event on Sunday 09 at 12:00, and a note on the same event stating the course is on October 10th at 4pm UK time. No event has been properly sent, and I receive an email confirmation of any participant that declines or accept the calendar event, spamming by inbox with 200> messages.
I still have not received the link to the zoom room, nor a proper calendar event invite.
Pim de Haan (Oct 10 2022 at 08:35):Hi everyone,
The first seminar is held today Monday October 10 on 4PM UK time (3PM UTC). The zoom link is https://uva-live.zoom.us/j/87524053828
See you then!
Pim
Petar Veličković (Oct 10 2022 at 09:45):Pim de Haan said:
Hi everyone,
The first seminar is held today Monday October 10 on 4PM UK time (3PM UTC). The zoom link is https://uva-live.zoom.us/j/87524053828
See you then!
Pim
Thanks Pim! I've just forwarded this to everyone on the Google Group. Hopefully it is visible :)
Jules Tsukahara (Oct 10 2022 at 12:26):Hi everyone, unfortunately I might not be able to make it to the first lecture. Will it be recorded?
Petar Veličković (Oct 10 2022 at 12:46):Jules Tsukahara said:
Hi everyone, unfortunately I might not be able to make it to the first lecture. Will it be recorded?
Hi Jules! Our hope is to record the lectures :)
Ross Gayler (Oct 10 2022 at 13:07):Petar Veličković said:
Jules Tsukahara said:
Hi everyone, unfortunately I might not be able to make it to the first lecture. Will it be recorded?
Hi Jules! Our hope is to record the lectures :)
I'm certainly hoping they can be recorded, because it's scheduled for 2am in my local time and I will be asleep.
Pim de Haan (Oct 10 2022 at 14:31):If everything works out, it'll be livestreamed to youtube at https://youtu.be/4poHENv4kR0 and permanently available from there.
Khalid Saifullah (Oct 10 2022 at 15:10):Hello everyone,
I can't find the zoom link for today's event, can someone please share? (sorry found it now!)
Debasish Ghosh (Oct 10 2022 at 16:07):Hello - Thanks for the great first lecture. Do we get new invites for each of the next lectures ?
Bruno Gavranović (Oct 10 2022 at 16:10):Hello everyone, thanks for the great attendance and interesting questions! I think we only got the chance to go through some questions , there were too many :sweat_smile:
Please do feel free to ask here.
Likewise - you might have noticed we didn't talk about any CT specifics in this lecture - this was intended. If you're hungry for more (and can't wait until the next week's lecture), here's the resources I promised to share with you:
Bruno Gavranović (Oct 10 2022 at 16:10):Debasish Ghosh said:
Hello - Thanks for the great first lecture. Do we get new invites for each of the next lectures ?
The calendar invite you've received should contain invites for next week as well
Debasish Ghosh (Oct 10 2022 at 16:13):Bruno Gavranovic said:
Debasish Ghosh said:
Hello - Thanks for the great first lecture. Do we get new invites for each of the next lectures ?
The calendar invite you've received should contain invites for next week as well
oops .. my bad. Should have checked before. Indeed they are there .. Thanks ..
Tom Schwaller (Oct 10 2022 at 16:13):Do you mind also sharing todays slide deck?
Tom Schwaller (Oct 10 2022 at 16:14):Studied Cellular sheaves and sheaf neural networks lately, so got quite excited when you mentioned sheaves today ;-)
Matteo Capucci (he/him) (Oct 10 2022 at 16:15):sheaves stan here
Bruno Gavranović (Oct 10 2022 at 16:20):Tom Schwaller said:
Do you mind also sharing todays slide deck?
Here are the slides
Nitin Kishore Sai Samala (Oct 10 2022 at 16:30):Newbie question : In today's lecture, we spoke about category theory being a bird's eye view of functors in groups, sets, vector spaces, networks, systems, etc., so it is something like "this is a set of different lego blocks and how some can connect with each other" or is it also possible to "build some compositional construct assuming unconstrained number of different lego blocks and their possible connections". Since we were talking about the combination of Geometric DL , I was wondering if there are any hands-on practical tutorials in this course that give a glimpse into this
Jules Tsukahara (Oct 10 2022 at 16:36):@Petar Veličković I think you mentioned the "naturality" of equivariant functions in the talk, is this related to natural transformations between functors and where can I read more about it?
Petar Veličković (Oct 10 2022 at 16:38):Jules Tsukahara said:
Petar Veličković I think you mentioned the "naturality" of equivariant functions in the talk, is this related to natural transformations between functors and where can I read more about it?
Exactly so @Jules Tsukahara! Equivariant functions (the essence of geometric deep learning) are a special case of natural transformations between functors.
The key paper exemplifying this is Natural Graph Networks (NeurIPS'20), and its first author @Pim de Haan will tell us all about it in Lecture 4 :)
Jules Tsukahara (Oct 10 2022 at 16:41):Thank you! Looking forward to it.
Emilio Minichiello (Oct 10 2022 at 17:24):Bruno Gavranovic said:
Hello everyone, thanks for the great attendance and interesting questions! I think we only got the chance to go through some questions , there were too many :sweat_smile:
Please do feel free to ask here.
Likewise - you might have noticed we didn't talk about any CT specifics in this lecture - this was intended. If you're hungry for more (and can't wait until the next week's lecture), here's the resources I promised to share with you:
Thanks for these references! As somebody with a background in pure math/category theory but little applied knowledge, do you have a recommendation for good references on where to get started with Machine Learning/Deep Learning?
Petar Veličković (Oct 10 2022 at 17:34):Emilio Minichiello said:
Bruno Gavranovic said:
Hello everyone, thanks for the great attendance and interesting questions! I think we only got the chance to go through some questions , there were too many :sweat_smile:
Please do feel free to ask here.
Likewise - you might have noticed we didn't talk about any CT specifics in this lecture - this was intended. If you're hungry for more (and can't wait until the next week's lecture), here's the resources I promised to share with you:Thanks for these references! As somebody with a background in pure math/category theory but little applied knowledge, do you have a recommendation for good references on where to get started with Machine Learning/Deep Learning?
Since you say you already have mathematical background, I would warmly recommend our proto-book on Geometric Deep Learning:
https://geometricdeeplearning.com/
which talks about how we can express basically all of deep learning in use today using equivariant maps.
Naturally, we'll explain in this course how Category Theory further extends these concepts. And in fact, in the full version of the GDL book (coming 2023) we will have a CT-based chapter.
The page I shared above has lots of different modalities of learning (book, blogs, talks, lecture course, webinar)... so you can mix and match however you prefer. :)
Emilio Minichiello (Oct 10 2022 at 17:55):Awesome, thank you!
Caio Oliveira (Oct 11 2022 at 02:24):(deleted)
Bruno Gavranović (Oct 11 2022 at 16:52):To add to answers of previous questions about what are good resources for learning CT, I forgot to recommend having a look at Math3ma's web page! It includes a lot of interesting drawings, ideas and approachable intros to CT.
Here are some blog posts that might be interesting:
ShivaKondapalli (Oct 15 2022 at 03:40):Thanks for the intro to cats and AI. My vague question: Is there any literature on using toposes and geometry for AI/DL. What promise does something like that hold for this field? I ask this because categories were in some sense developed for (algebraic) topology and geometry, so can we use constructions in those fields for the benefit of AI/ML?
Matteo Capucci (he/him) (Oct 16 2022 at 00:52):This paper does exactly that, but it's quite infamous in the community for being at least unclear on what exactly they are doing with topoi.
Matteo Capucci (he/him) (Oct 16 2022 at 00:53):I'm sure the current flurry of activity on sheaves for GNNs (and the advent of geometric deep learning) will eventually summon toposic tooling, though I don't exactly understand how atm
Matteo Capucci (he/him) (Oct 16 2022 at 00:57):Topoi are a great thing but they can be rather abstract and unwiedy, so often they are used to mint general results rather than directly in applications. I suspect we are going to see topos-inspired machinery being deployed to ML (e.g. various flavours of sheaf cohomology) way before we see topos theory being directly used to do ML.
Abraham Btesh (Oct 16 2022 at 16:20):Does anybody know where I can find the link to the recording of this lecture?
Petar Veličković (Oct 16 2022 at 19:38):Pim de Haan said:
If everything works out, it'll be livestreamed to youtube at https://youtu.be/4poHENv4kR0 and permanently available from there.
@Abraham Btesh Pim shared it here :)
Marco Selvi (Oct 17 2022 at 13:58):Does anyone have a link to the slides used in the first lecture that they could share? (Apologies if it was already posted and I missed it in emails / previous messages)
Pim de Haan (Oct 17 2022 at 15:09):Slides at http://cats.for.ai/program/