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.
Pim de Haan (Nov 18 2022 at 11:35):Hi everyone!
The second guest lecture will be given by @Taco Cohen, titled: "Causal Model Abstraction & Grounding via Category Theory"
The abstract for Taco's lecture is as follows:
Causal models are used in many areas of science to describe data generating processes and reason about the effect of changes to these processes (interventions). Causal models are typically highly abstracted representations of the underlying process, consisting of only a few carefully selected variables, and the causal mechanisms between them. This simplifies causal reasoning, but the relation between the model and the underlying system is never described in mathematical terms, and this has led to considerable philosophical confusions. Furthermore, it has made it hard to understand how causal modeling relates to other fields such as physics (where systems are described by dynamical laws without reference to causes), dynamical systems, and agent-centric frameworks such as Markov Decision Processes (MDPs). In this talk we study this idea of abstraction from a categorical perspective, focussing on two questions in particular:
Among other things, this guest lecture will help explain key parts of Taco's paper "Towards a Grounded Theory of Causation for Embodied AI" https://arxiv.org/abs/2206.13973
Taco's guest lecture will take place in the usual slot, next week (Monday 21 November, starting 4PM UK Time). The lecture will be given on Zoom and live-streamed on YouTube, just as before.
Zoom: https://uva-live.zoom.us/j/83816139841
YouTube: https://youtu.be/5mZhcXhbciE
Thanks so much to those that have already filled in the feedback form. For those that haven't yet: you'd help us a lot by taking a few minutes to share you thoughts on the course on this form https://forms.gle/gRpv1t9Xi6zzpyCB7 .
Bruno Gavranović (Nov 21 2022 at 10:40):I'm getting messages from people that this event still isn't showing up in their calendar. There could be a variety of reasons, the most likely one being that the calendar was imported using the .ical. file from the original email (which doesn't allow us to push updates to the calendar), and not the link.
I just created a new calendar link, and checked using a separate account that pushing updates works on this one. So if you've had any issues, this should work, and it should correctly add future events. But if everything was already working for you, then you don't need to change anything, as the underlying calendar is the same.
Bruno Gavranović (Nov 21 2022 at 10:43):Apologies for this error - we should've foreseen that calendar imports using the .ical file don't allow for updates.
Ieva Cepaite (Nov 21 2022 at 11:03):Bruno Gavranovic said:
I'm getting messages from people that this event still isn't showing up in their calendar. There could be a variety of reasons, the most likely one being that the calendar was imported using the
.ical.file from the original email (which doesn't allow us to push updates to the calendar), and not the link.I just created a new calendar link, and checked using a separate account that pushing updates works on this one. So if you've had any issues, this should work, and it should correctly add future events. But if everything was already working for you, then you don't need to change anything, as the underlying calendar is the same.
This works great now, thanks Bruno!
Jules Tsukahara (Nov 21 2022 at 20:24):Thank you for a really interesting talk! During the whole talk I couldn't help but think of Probabilistic Graphical Models, yet it wasn't mentioned once. I have the feeling that PGMs are less expressive than the SCM framework, but it got me wondering whether those two fields talk to each other or not.
Taco Cohen (Nov 21 2022 at 20:51):Jules Tsukahara said:
Thank you for a really interesting talk! During the whole talk I couldn't help but think of Probabilistic Graphical Models, yet it wasn't mentioned once. I have the feeling that PGMs are less expressive than the SCM framework, but it got me wondering whether those two fields talk to each other or not.
Oh yes, they're very much related. PGMs are basically the same as what I called Bayesian Networks. However, it is often the case that one can model a given distribution with a PGM/BN, but one can NOT expect such model to make correct predictions for interventions (i.e. when you modify the PGM by removing some edge).
Jules Tsukahara (Nov 22 2022 at 18:24):Taco Cohen said:
Oh yes, they're very much related. PGMs are basically the same as what I called Bayesian Networks. However, it is often the case that one can model a given distribution with a PGM/BN, but one can NOT expect such model to make correct predictions for interventions (i.e. when you modify the PGM by removing some edge).
I see, thank you. PGMs seem to be the basic object of study of Causal Inference, but additional rules such as interventions must be taken into account to capture SCMs. Would it be possible to get the slides for your talk, if they haven't been shared already?
Pim de Haan (Nov 23 2022 at 17:49):The slides can be found at https://cats.for.ai/program/
Jules Tsukahara (Nov 23 2022 at 19:43):Oh I missed them sorry, thanks a lot!