Forman--Ricci Curvature on Contact-Sequence Temporal Networks via Spatiotemporal Prism Complexes

TL;DR

This paper introduces a geometric framework using Forman--Ricci curvature on contact-sequence temporal networks via spatiotemporal prism complexes, enabling detailed analysis of dynamic interaction data.

Contribution

It develops a novel simplicial complex construction for temporal networks and compares two variants of Forman--Ricci curvature, providing theoretical insights and practical tools.

Findings

The two Forman variants disagree on 56-67% of 1-simplices in experiments.

The variants are strongly correlated despite disagreements.

The framework offers a parameter-free method for curvature assignment in temporal data.

Abstract

Temporal networks -- sequences of time-stamped contacts among nodes -- constitute the finest-grained representation of dynamic interaction data; however, geometric and topological analyses of such networks have remained largely confined to time-aggregated or snapshot-based approximations. Such reductions destroy the temporal ordering and interevent statistics essential for understanding spreading dynamics, synchronization, and information flow. This study proposes a geometric framework that lifts a contact-sequence temporal network into a genuine simplicial complex through a prism construction adapted from algebraic topology. On this spatiotemporal prism complex, we develop the Forman--Ricci curvature in its original CW-complex form and contrast it with an augmented variant widely used in network science. We prove that the two variants coincide under uniform weights, derive a closed-form expression for their pointwise discrepancy in the general case, and identify the precise conditions under which they diverge -- conditions generically satisfied in temporal networks because temporal edges carry interval-dependent weights. Numerical experiments on three synthetic contact-network models (Erd\H os--R\'enyi, activity-driven, and bursty) and on the SocioPatterns Hypertext 2009 face-to-face contact dataset quantitatively confirm the theoretical predictions: the two Forman variants disagree on $56$--$67\%$ of the $1$-simplices -- predominantly the temporal and diagonal simplices -- while remaining strongly correlated according to the Pearson coefficient. The proposed framework provides a principled, parameter-free method for assigning discrete Ricci curvature to each contact event, thereby opening a new geometric avenue for temporal data analysis.