Adaptive Graph Pruning with Sudden-Events Evaluation for Traffic Prediction using Online Semi-Decentralized ST-GNNs

IVAN KRALJ 1, LODOVICO GIARETTA 2, GORDAN JEŽIĆ 1, IVANA PODNAR ŽARKO 1, ŠARŪNAS GIRDZIJAUSKAS 2 3 · 2025

Read Adaptive Graph Pruning with Sudden-Events Evaluation for Traffic Prediction using Online Semi-Decentralized ST-GNNs by IVAN KRALJ 1, LODOVICO GIARETTA 2, GORDAN JEŽIĆ 1, IVANA PODNAR ŽARKO 1, ŠARŪNAS GIRDZIJAUSKAS 2 3 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-17352.txt.

Description

Ivan Kralj and four co-authors trim the chatter between roadside computers that forecast traffic together. Their forecast models read streams from sensors spread over a city, split across small edge servers called cloudlets. Neighboring servers keep resending each other the same sensor data, which wastes bandwidth and slows the whole system down.

Their fix prunes the shared data on the fly, cutting what recent performance shows is not helping, while guarding the links that matter. They also coin a new score, SEPA, that grades how fast a model reacts to sudden jams and clear-ups. On two large California traffic data sets, pruning kept forecasts sharp while sending far less data. The new score revealed links that plain error scores had undervalued.

Catalog record

CREATORIVAN KRALJ 1, LODOVICO GIARETTA 2, GORDAN JEŽIĆ 1, IVANA PODNAR ŽARKO 1, ŠARŪNAS GIRDZIJAUSKAS 2 3
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE79 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2512-17352

Browse the full library · About the Tower of Babel