4/5 We show how to factorize this equivalence distribution using latent categorical variables—a kind of embeddings.
Plus we develop exact (via singularization + magic sets) and approximate (Monte Carlo) inference methods.
3/5 Instead of a distribution over possible worlds (standard probabilistic logic), we define a distribution over equivalence relations between symbols.
2/5 We show that soft unification, which combines neural embeddings with symbolic rules, suffers from local optima and leaks probability mass. Our fix: equivalence semantics.
1/5 Tomorrow I’ll talk about the 𝐩𝐫𝐨𝐛𝐚𝐛𝐢𝐥𝐢𝐬𝐭𝐢𝐜 𝐩𝐫𝐨𝐠𝐫𝐚𝐦𝐦𝐢𝐧𝐠 𝐬𝐞𝐦𝐚𝐧𝐭𝐢𝐜𝐬 𝐨𝐟 𝐝𝐢𝐟𝐟𝐞𝐫𝐞𝐧𝐭𝐢𝐚𝐛𝐥𝐞 𝐩𝐫𝐨𝐯𝐢𝐧𝐠 at #NeurIPS San Diego (poster #614 11am).
📃 openreview.net/pdf?id=rEUbD...
📺 www.youtube.com/watch?v=sOTX...
Scholar Inbox for NeurIPS is live now.
Terence Tao (@teorth.bsky.social) has written a thread on Mastodon about the impact of the federal grant freeze to UCLA, particularly to his own field of Mathematics. UCLA's IPAM (Institute of Pure and Applied Mathematics) could shut down entirely
mathstodon.xyz/@tao/1149568...
We’re proud to announce the launch of AutumnBench, an open-source benchmark developed on our Autumn platform. This benchmark, led by our MARA team, provides a novel platform for evaluating world modeling and causal reasoning in both human and artificial intelligence.
The crazy thing about epicycles: at the time Kopernicus introduced his model they were far superior in terms of predictive power compared to the heliocentric model. This was still true when Kepler refined the model with ellipses. Epicycles had been extremely refined over time (capital+labor)
We propose Neurosymbolic Diffusion Models! We find diffusion is especially compelling for neurosymbolic approaches, combining powerful multimodal understanding with symbolic reasoning 🚀
Read more 👇
Just under 10 days left to submit your latest endeavours in #tractable probabilistic models!
Join us at TPM @auai.org #UAI2025 and show how to build #neurosymbolic / #probabilistic AI that is both fast and trustworthy!
This was joint work with my amazing co-authors @pedrozudo.bsky.social and @vincentderk.bsky.social
We developed a library to make logical reasoning embarrasingly parallel on the GPU.
For those at ICLR 🇸🇬: you can get the juicy details tomorrow (poster #414 at 15:00). Hope to see you there!
And @gabventurato.bsky.social who presented on 3 different days!
We generalize backpropgation to semirings, which allows us to unify many learning tasks. This is similar to how inference on probabilistic models has been generalised (e.g. sum-product and max-product algorithms)
We all know backpropagation can calculate gradients, but it can do much more than that!
Come to my #AAAI2025 oral tomorrow (11:45, Room 119B) to learn more.
Are you at AAAI in Philadelphia and interested about #tensor-factorizations or #circuits or even both?
Then join us today at our tutorial: "From tensor factorizations to circuits (and back!)"
Details and materials here
april-tools.github.io/aaai25-tf-pc...
Time 4:15pm - 6:00pm, Room 117
Congratulations!
🔥 Can AI reason over time while following logical rules in relational domains? We will present Relational Neurosymbolic Markov Models (NeSy-MMs) next week at #AAAI2025! 🎉
📜 Paper: arxiv.org/pdf/2412.13023
💻 Code: github.com/ML-KULeuven/...
🧵⬇️
I'm visiting the StarAI lab of @guyvdb.bsky.social at UCLA for a couple of months starting this week.
If you're around and want to have a chat let me know :)
Happy to see our work at TMLR!
We systematically show the relationships between two apparently different fields: tensor factorizations and circuits, and how bridging the two enables us to exchange results, research opportunitie in ML, and practical implementation solutions.
We are hiring for PhD positions! I’m looking for people interested in exploring the intersection of learning and reasoning with applications to anomaly detection and sports. @dtai-kuleuven.bsky.social @wannesm.bsky.social
www.kuleuven.be/personeel/jo...
Unsure where to submit your next research paper to now that aideadlin.es is not updated anymore? And let’s be honest, is the location not as important as the conference itself?
🗺️ Check out my latest side-project: deadlines.pieter.ai
Are you interested in more scalable reasoning under uncertainty and attending NeurIPS? Then pass by our poster #3708 later today at 4.30pm! 🕟
We use recursive integer arithmetic to express combinatorial problems and add uncertainty. Inference can be massively accelerated with tensors and the FFT. 🚀
🚨 Interpretable AI often means sacrificing accuracy—but what if we could have both? Most interpretable AI models, like Concept Bottleneck Models, force us to trade accuracy for interpretability.
But not anymore, due to Concept-Based Memory Reasoner (CMR)! #NeurIPS2024 (1/7)
You can also do it without independence, but instead using distributivity, cf. arithmetic circuits.
