DHNs with max/sum/mean aggregations correspond to UNFO and extensions with counting quantifiers, enabling decidability analysis for emptiness and subsumption problems.
Unifying approach to uniform expressivity of graph neural networks
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
years
2026 2verdicts
UNVERDICTED 2representative citing papers
Rational-weight ReLU neural networks and generalized polynomial rings are characterized as formulas in RPL, RPL(⊙)≤1, and fragments of LΠ1/2 logic.
citing papers explorer
-
Expressive Power of Deep Homomorphism Networks over Relational Databases
DHNs with max/sum/mean aggregations correspond to UNFO and extensions with counting quantifiers, enabling decidability analysis for emptiness and subsumption problems.
-
Neural networks as fuzzy logic formulas
Rational-weight ReLU neural networks and generalized polynomial rings are characterized as formulas in RPL, RPL(⊙)≤1, and fragments of LΠ1/2 logic.