Indep NeSy goes to ICML 2024

april-tools · May 3, 2024 · 19b463b · 19b463b
1 parent 5cb129b
commit 19b463b
Show file tree

Hide file tree

Showing 2 changed files with 9 additions and 2 deletions.
diff --git a/_news/indep2024icml.md b/_news/indep2024icml.md
@@ -0,0 +1,7 @@
+---
+title: "Independence Assumption NeSy accepted at ICML 2024"
+collection: news
+permalink: /news/indep-nesy-icml2024
+date: 2024-05-01
+---
+Our work on <a href="https://arxiv.org/abs/2404.08458"><b>what common assumptions in neurosymbolic systems imply in terms of expressiveness and learnability</b></a> is accepted at ICML 2024.
diff --git a/_publications/vankrieken2024indep.md b/_publications/vankrieken2024indep.md
@@ -9,7 +9,7 @@ image: "/images/papers/vankrieken2024indep/indep.png"
 authors: "Emile van Krieken, Pasquale Minervini, Edoardo M. Ponti, Antonio Vergari"
 paperurl: "https://arxiv.org/abs/2404.08458"
 pdf: "https://arxiv.org/pdf/2404.08458.pdf"
-venue: "arXiv 2024"
+venue: "ICML 2024"
 excerpt: "We theoretically analyze the common assumption that many NeSy models -- from the semantic loss to deep Problog -- do: the independence among terms of a logical formula, and highlight how this biases learning and make some solutions impossible to retrieve."
 abstract: "State-of-the-art neurosymbolic learning systems use probabilistic reasoning to guide neural networks towards predictions that conform to logical constraints over symbols. Many such systems assume that the probabilities of the considered symbols are conditionally independent given the input to simplify learning and reasoning. We study and criticise this assumption, highlighting how it can hinder optimisation and prevent uncertainty quantification. We prove that loss functions bias conditionally independent neural networks to become overconfident in their predictions. As a result, they are unable to represent uncertainty over multiple valid options. Furthermore, we prove that these loss functions are difficult to optimise: they are non-convex, and their minima are usually highly disconnected. Our theoretical analysis gives the foundation for replacing the conditional independence assumption and designing more expressive neurosymbolic probabilistic models."
 supplemental: 
@@ -19,4 +19,4 @@ bibtex: "@article{vankrieken2024indep,<br/>
   journal={arXiv preprint arXiv:404.08458},<br/>
   year={2024}
 }"
----
+---