Saved in:
Bibliographic Details
Main Authors: Panyshev, Alexander, Vinichenko, Dmitry, Travkin, Oleg, Alferov, Roman, Zaytsev, Alexey
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2505.12526
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912381396844544
author Panyshev, Alexander
Vinichenko, Dmitry
Travkin, Oleg
Alferov, Roman
Zaytsev, Alexey
author_facet Panyshev, Alexander
Vinichenko, Dmitry
Travkin, Oleg
Alferov, Roman
Zaytsev, Alexey
contents Temporal Graph Networks (TGNs), while being accurate, face significant training inefficiencies due to irregular supervision signals in dynamic graphs, which induce sparse gradient updates. We first theoretically establish that aggregating historical node interactions into pseudo-labels reduces gradient variance, accelerating convergence. Building on this analysis, we propose History-Averaged Labels (HAL), a method that dynamically enriches training batches with pseudo-targets derived from historical label distributions. HAL ensures continuous parameter updates without architectural modifications by converting idle computation into productive learning steps. Experiments on the Temporal Graph Benchmark (TGB) validate our findings and an assumption about slow change of user preferences: HAL accelerates TGNv2 training by up to 15x while maintaining competitive performance. Thus, this work offers an efficient, lightweight, architecture-agnostic, and theoretically motivated solution to label sparsity in temporal graph learning.
format Preprint
id arxiv_https___arxiv_org_abs_2505_12526
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Never Skip a Batch: Continuous Training of Temporal GNNs via Adaptive Pseudo-Supervision
Panyshev, Alexander
Vinichenko, Dmitry
Travkin, Oleg
Alferov, Roman
Zaytsev, Alexey
Machine Learning
Temporal Graph Networks (TGNs), while being accurate, face significant training inefficiencies due to irregular supervision signals in dynamic graphs, which induce sparse gradient updates. We first theoretically establish that aggregating historical node interactions into pseudo-labels reduces gradient variance, accelerating convergence. Building on this analysis, we propose History-Averaged Labels (HAL), a method that dynamically enriches training batches with pseudo-targets derived from historical label distributions. HAL ensures continuous parameter updates without architectural modifications by converting idle computation into productive learning steps. Experiments on the Temporal Graph Benchmark (TGB) validate our findings and an assumption about slow change of user preferences: HAL accelerates TGNv2 training by up to 15x while maintaining competitive performance. Thus, this work offers an efficient, lightweight, architecture-agnostic, and theoretically motivated solution to label sparsity in temporal graph learning.
title Never Skip a Batch: Continuous Training of Temporal GNNs via Adaptive Pseudo-Supervision
topic Machine Learning
url https://arxiv.org/abs/2505.12526