activity
20242026
collaborators

5 papers

cs.LG2026

Learning to Execute Graph Algorithms Exactly with Graph Neural Networks

Muhammad Fetrat Qharabagh, Artur Back de Luca, George Giapitzakis +1

Understanding what graph neural networks can learn, especially their ability to learn to execute algorithms, remains a central theoretical challenge. In this work, we prove exact l…

cs.LG2026

Certification from Examples is Hard for Circuits and Transformers under Minimal Overparametrization

Artur Back de Luca, Kimon Fountoulakis

As state-of-the-art neural networks are deployed on reasoning and algorithmic tasks, exactness guarantees become increasingly important. However, high average-case accuracy can sti…

cs.LG2026

Learning to Add, Multiply, and Execute Algorithmic Instructions Exactly with Neural Networks

Artur Back de Luca, George Giapitzakis, Kimon Fountoulakis

Neural networks are known for their ability to approximate smooth functions, yet they fail to generalize perfectly to unseen inputs when trained on discrete operations. Such operat…

cs.LG2025

Positional Attention: Expressivity and Learnability of Algorithmic Computation

Artur Back de Luca, George Giapitzakis, Shenghao Yang +2

There is a growing interest in the ability of neural networks to execute algorithmic tasks (e.g., arithmetic, summary statistics, and sorting). The goal of this work is to better u…

cs.LG2024

Simulation of Graph Algorithms with Looped Transformers

Artur Back de Luca, Kimon Fountoulakis

The execution of graph algorithms using neural networks has recently attracted significant interest due to promising empirical progress. This motivates further understanding of how…