What makes a good feedforward computational graph?

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Vitvitskyi, Alex, Araújo, João G. M., Lackenby, Marc, Veličković, Petar
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866908387440066560
author Vitvitskyi, Alex
Araújo, João G. M.
Lackenby, Marc
Veličković, Petar
author_facet Vitvitskyi, Alex
Araújo, João G. M.
Lackenby, Marc
Veličković, Petar
contents As implied by the plethora of literature on graph rewiring, the choice of computational graph employed by a neural network can make a significant impact on its downstream performance. Certain effects related to the computational graph, such as under-reaching and over-squashing, may even render the model incapable of learning certain functions. Most of these effects have only been thoroughly studied in the domain of undirected graphs; however, recent years have seen a significant rise in interest in feedforward computational graphs: directed graphs without any back edges. In this paper, we study the desirable properties of a feedforward computational graph, discovering two important complementary measures: fidelity and mixing time, and evaluating a few popular choices of graphs through the lens of these measures. Our study is backed by both theoretical analyses of the metrics' asymptotic behaviour for various graphs, as well as correlating these metrics to the performance of trained neural network models using the corresponding graphs.
format Preprint
id arxiv_https___arxiv_org_abs_2502_06751
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle What makes a good feedforward computational graph?
Vitvitskyi, Alex
Araújo, João G. M.
Lackenby, Marc
Veličković, Petar
Machine Learning
Artificial Intelligence
Social and Information Networks
As implied by the plethora of literature on graph rewiring, the choice of computational graph employed by a neural network can make a significant impact on its downstream performance. Certain effects related to the computational graph, such as under-reaching and over-squashing, may even render the model incapable of learning certain functions. Most of these effects have only been thoroughly studied in the domain of undirected graphs; however, recent years have seen a significant rise in interest in feedforward computational graphs: directed graphs without any back edges. In this paper, we study the desirable properties of a feedforward computational graph, discovering two important complementary measures: fidelity and mixing time, and evaluating a few popular choices of graphs through the lens of these measures. Our study is backed by both theoretical analyses of the metrics' asymptotic behaviour for various graphs, as well as correlating these metrics to the performance of trained neural network models using the corresponding graphs.
title What makes a good feedforward computational graph?
topic Machine Learning
Artificial Intelligence
Social and Information Networks
url https://arxiv.org/abs/2502.06751