敬邀參加 IDA 智慧設計自動化聯盟國外專家講座！Welcome to join this talk!
Title: Reasoning on Graph Neural Networks (GNN) via Logical Characterisation
Speaker:
Dr. Tony Tan, University of Liverpool
Date & Time:
14:00-15:00, Friday, July 10, 2026
Location:
Room 145, EE2 Building (電機二館), National Taiwan University
Abstract:
Graph Neural Networks (GNN) are neural network architecture for machine learning problems on graph.
In this talk we will give an overview of the connection between GNN and the classical logics. We will start with the classical result which states that the expressiveness of GNNs can be characterised precisely by the combinatorial Weisfeiler-Leman algorithms and by finite variable counting first-order logics. Building on this, we will present recent results concerning the expressiveness and decidability of a popular GNN formalism, exploiting connections with logic, in particular with recently-discovered decidable logics involving “Presburger quantifiers”. In most cases, these logics can be used to measure the expressiveness of classes of GNNs, in some cases getting exact correspondences between the expressiveness of logics and GNNs.
If time permits, we will also briefly touch on the robustness problem for GNN.We will make this talk non-technical as much as possible, focussing more on the intuition and survey of recent results.
Bio:
Dr. Tony Tan is a senior lecturer (associate professor) in the School of Computer Science and Informatics in the University of Liverpool, UK, where he currently serves as the director of the postgraduate teaching programme. Prior to that, he spent 8 wonderful years in the department of Computer Science and Information Engineering in National Taiwan University, first as an assistant professor and later as associate professor.
His current research focus is logic and automated reasoning with applications on succinct SAT, GNN reasoning and circuit design.
報名網址：https://ppt.cc/fRHWlx
