Generating Temporal Networks with the Ascona Model

Samuel Koovely · 2025

Read Generating Temporal Networks with the Ascona Model by Samuel Koovely (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-2512-16972.txt.

Introduction

Samuel Koovely introduces the Ascona model, a queueing-based way to build continuous-time temporal networks, where links switch on and off over time. Such models test tools that hunt for communities, scale patterns, change points, and cycles in network data.

In the Markovian version, link start times follow a Poisson process and link spans are exponentially spread. Koovely derives traits of the link streams and uses them to make synthetic networks with set smoothness and event patterns. Joining the time mechanism to block-structured endpoint rules gives a continuous-time cousin of stochastic block models. Extensions reach discrete time and instant-contact limits.

Catalog record

CREATORSamuel Koovely
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE42 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2512-16972

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Generating Temporal Networks with the Ascona Model Samuel Koovely Abstract We introduce a new sampling method for continuous-time temporal networks based on queueing processes. In particular, we focus on a Markovian version of the model where the links between nodes are Poisson distributed in time and have exponential duration. We highlight the stochastic properties of these temporal structures and leverage them to design synthetic temporal networks with a controllable level of smoothness, which follow patterns relevant for the validation and interpretation of methods for community, scale, change-point, and periodicity detection. Additionally, we show that imposing assortativity constraints on the samples leads to a continuous-time generalization of stochastic block models. Finally, we describe how variations of the model can be used to sample alternative types of structure and temporal networks, especially discrete-time ones. 1 Introduction Temporal (or dynamic) networks/graphs are extensions of graphs that incorporate temporal information. Various mathematical objects can be considered to be temporal networks, and different formalisms have been proposed to describe them (e.g., [ 8 , 17 , 20 ] ). These objects can be classified on a spectrum of increasing complexity based on the granularity of the encoded temporal information [ 29 , 28 ] . On one side, we have static graphs, which aggregate in one structure all temporal information; on the other side, there are stream graphs and link streams that avoid any aggregation. Somewhere in between the two sides lie snapshot networks and discrete-time stream graphs/link streams. Temporal networks are relevant in modeling many complex systems [ 24 ] , and, as for their static counterparts, understanding their communities/clusters is a fundamental problem in network science. The community structure evolution is typically characterized by archetypal structural changes [ 10 ] , called events , or combinations thereof [ 14 ] . Therefore, sampling methods that allow to model these archetypes are of paramount importance for the generation of quality synthetic data with controlled ground-truth. One needs these synthetic networks, for instance, for the validation and interpretation of algorithms tasked for community, scale, change-point, and periodicity detection. A review of the main existing methods, particularly relevant for community detection benchmarks, can be found in [ 9 ] . For some applications, one would also want to sample networks that are less artificial and instead are closer to real-world datasets. Recently proposed techniques in this direction (e.g., [ 15 , 23 , 26 ] ) aim at producing surrogate temporal graphs, structures that maintain some or most properties of a reference (often real-world) temporal network. However, these methods do not provide sufficient control to design a controlled experimental setup, as they bypass the explicit formulation of a modeling mechanism by relying on a pre-existing structure. Agent-based models (e.g., [ 25 , 30 ] ) are an alternative option. They leverage control of microscopic dynamics of a complex system and combine them so that, in the aggregate, complex phenomena emerge. However, often, the understanding of these models relies heavily on simulations, as the underlying dynamics are complex, leading to mathematical intractability, hence limited stochastic and statistical control. Most of the temporal network literature, especially on sampling methods, has focused on snapshot networks or discrete-time structures, leaving out the more complex continuous-time ones (a notable exception being [ 6 ] ). The main contribution of this article is the introduction of a sampling method for continuous-time temporal networks, and the explanation of how to use it to sample data reflecting the above-mentioned event patterns, plus a continuous-time generalization of the stochastic block model (SBM) [ 1 ] . The model works in two fundamental steps: in the first, we sample when nodes are connected with a link, and in the second, we sample which nodes to assign to each link sampled in the previous step. Separating the temporal from the connectivity aspect offers benefits in terms of implementation and clarity, while not sacrificing expressivity. This separation mechanism also helps to structure the mathematical analysis of a temporal network (model), as the two aspects can be considered orthogonal to each other, hence treated separately, as in [ 4 ] . Connectivity can exhibit almost arbitrary levels of complexity and is responsible for various structural aspects such as assortativity, disassortativity, and nestedness. A complete analysis of its properties from a stochastic and statistical perspective is therefore challenging. The temporal aspect appears to be simpler; however, surprisingly, its probabilistic and statistical analysis has been neglected. This work addresses this gap: its key idea is to describe the appearance and disappearance of nodes’ connections in terms of queueing processes, and then focus on the evolution of the total number of active connections as a function of time to derive probabilistic and statistical results. Queueing processes were originally introduced to describe the flow of customers (jobs, particles, calls) arriving at a service facility. It is possible to approximate the evolution of most queueing systems with Markov chains. The advantage of these is that it is possible to calculate the stationary distribution of a chain by solving a linear system of equations called the balance conditions [ 32 ] . The structure of this work is as follows. In Section 2 we introduce the notation and describe the objects studied in the article. It is divided into two parts: the first is devoted to formalisms of temporal networks, the second to queueing processes. In Section 3 we introduce the Ascona model, a temporal networks model based on queueing theory, and then use a Markovian parametrization of it to generate link streams with desired temporal evolutions. Based on this Markovian framework, we describe a link stream generalization of SBMs, and compare it to their currently most popular temporal extensions, dynamic SBMs. We conclude the section with a description of variations of the Ascona model. The end of the article, Section 4 , contains a discussion of the presented results and highlights possible future research directions stemming from this work. 2 Theoretical Framework In this section, we describe the notation and objects we need to introduce the Ascona model. In the first part, we describe various formalisms of temporal networks following the framework presented in [ 20 ] , and in the second part, general queueing processes, focusing on the M/M/\infty queue, also known as the ample-server model [ 16 , Ch. 2.7] . 2.1 Temporal Networks We model a dynamic network within a time domain T (either an interval of \mathbb{R} for continuous time or a finite subset of \mathbb{Z} for discrete time) and over a finite node set V . In the most general setup, stream graphs, V is not fixed in time. However, for many, if not most applications, a disappearing node is equivalent to a node with zero degree that never disappears. Therefore, in Section 3 , we focus on the goal of producing a sampling method for link streams, where V is constant. In Section 4 , we are going to briefly discuss how one could extend the model to generate stream graphs. 2.1.1 Link Streams The link stream is a temporal network formalism that concentrates on the timing of interactions. There are two common variants. General Model A link stream is a triplet L=(T,V,E),\qquad E\subseteq\bigl\{[\alpha,\omega)\times\{\{u,v\}\}:\alpha\leqslant\omega,u\neq v\bigr\}, (2.1) so that

Read the complete text