VC dimension of Graph Neural Networks with Pfaffian activation functions

Download PDF
Open Webpage

Giuseppe Alessio D'Inverno, Monica Bianchini, Franco Scarselli

Published: 01 Jan 2025, Last Modified: 01 Aug 2025Neural Networks 2025EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•Establishment of bounds on the VC dimension for modern GNNs with Pfaffian activations•Bounds provided w.r.t. network parameters and number of colors deriving from WL test•An experimental validation of the theoretical analysis is conducted