Adaptive Memory Retention in Dynamic Graphs

Fabrizio De Castelli, Alessio Gravina, Moshe Eliasof, Carola-Bibiane Schönlieb, Davide Bacciu
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:23292-23315, 2026.

Abstract

Modeling graphs demands a careful balance between long-range propagation of information across nodes and the controlled dissipation of noisy or redundant signals to ensure stable learning and generalization. This challenge is exacerbated in dynamic graphs, where structural and temporal information interact, leading to uncontrolled information accumulation and amplifying noise, thereby affecting generalization. We introduce LAMP, a dynamic graph model for snapshot-based dynamic graphs that incorporates adaptive, learned dissipation within a principled dynamical systems framework. Our architecture combines impulsive neural ODEs with an antisymmetric parameterization to model conservative information flow, alongside data-driven dissipative dynamics that regulate information retention over space and time. This formulation yields stable yet expressive representations and enables effective long-range dependency modeling while avoiding pathological information buildup. We provide a theoretical analysis establishing stability guarantees and characterizing the representational power. Extensive experiments on synthetic and real-world benchmarks demonstrate state-of-the-art performance, particularly on tasks requiring extended-range dependency modeling.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-de-castelli26a, title = {Adaptive Memory Retention in Dynamic Graphs}, author = {De Castelli, Fabrizio and Gravina, Alessio and Eliasof, Moshe and Sch\"{o}nlieb, Carola-Bibiane and Bacciu, Davide}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {23292--23315}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/de-castelli26a/de-castelli26a.pdf}, url = {https://proceedings.mlr.press/v306/de-castelli26a.html}, abstract = {Modeling graphs demands a careful balance between long-range propagation of information across nodes and the controlled dissipation of noisy or redundant signals to ensure stable learning and generalization. This challenge is exacerbated in dynamic graphs, where structural and temporal information interact, leading to uncontrolled information accumulation and amplifying noise, thereby affecting generalization. We introduce LAMP, a dynamic graph model for snapshot-based dynamic graphs that incorporates adaptive, learned dissipation within a principled dynamical systems framework. Our architecture combines impulsive neural ODEs with an antisymmetric parameterization to model conservative information flow, alongside data-driven dissipative dynamics that regulate information retention over space and time. This formulation yields stable yet expressive representations and enables effective long-range dependency modeling while avoiding pathological information buildup. We provide a theoretical analysis establishing stability guarantees and characterizing the representational power. Extensive experiments on synthetic and real-world benchmarks demonstrate state-of-the-art performance, particularly on tasks requiring extended-range dependency modeling.} }
Endnote
%0 Conference Paper %T Adaptive Memory Retention in Dynamic Graphs %A Fabrizio De Castelli %A Alessio Gravina %A Moshe Eliasof %A Carola-Bibiane Schönlieb %A Davide Bacciu %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-de-castelli26a %I PMLR %P 23292--23315 %U https://proceedings.mlr.press/v306/de-castelli26a.html %V 306 %X Modeling graphs demands a careful balance between long-range propagation of information across nodes and the controlled dissipation of noisy or redundant signals to ensure stable learning and generalization. This challenge is exacerbated in dynamic graphs, where structural and temporal information interact, leading to uncontrolled information accumulation and amplifying noise, thereby affecting generalization. We introduce LAMP, a dynamic graph model for snapshot-based dynamic graphs that incorporates adaptive, learned dissipation within a principled dynamical systems framework. Our architecture combines impulsive neural ODEs with an antisymmetric parameterization to model conservative information flow, alongside data-driven dissipative dynamics that regulate information retention over space and time. This formulation yields stable yet expressive representations and enables effective long-range dependency modeling while avoiding pathological information buildup. We provide a theoretical analysis establishing stability guarantees and characterizing the representational power. Extensive experiments on synthetic and real-world benchmarks demonstrate state-of-the-art performance, particularly on tasks requiring extended-range dependency modeling.
APA
De Castelli, F., Gravina, A., Eliasof, M., Schönlieb, C. & Bacciu, D.. (2026). Adaptive Memory Retention in Dynamic Graphs. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:23292-23315 Available from https://proceedings.mlr.press/v306/de-castelli26a.html.

Related Material