Local equations for the generalized Lotka-Volterra model on sparse asymmetric graphs

D. Machado, P. Valigi, T. Tonolo, M. C. Angelini · 2025

Read Local equations for the generalized Lotka-Volterra model on sparse asymmetric graphs by D. Machado, P. Valigi, T. Tonolo, M. C. Angelini (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-2511-17499.txt.

Introduction

David Machado and three co-authors develop a method for predicting the fate of ecosystems modeled on sparse, asymmetric interaction networks. Real food webs and microbial communities are exactly that: most species pairs never interact, and influence rarely runs equally both ways. The generalized Lotka-Volterra model captures such communities, but its stochastic dynamics on sparse graphs resisted standard analysis. The authors derive local Fokker-Planck equations for each species and close the system with a mean-field approximation.

Stationary states then become cheap to compute for both symmetric and asymmetric interactions. The team validates the method against direct integration of the dynamics and against known results, and produces the first phase diagram for sparse asymmetric networks. The framework also transfers to other fields with similar network dynamics, such as economics and evolutionary game theory.

Catalog record

CREATORD. Machado, P. Valigi, T. Tonolo, M. C. Angelini
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE101 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-17499

Browse the full library · About the Tower of Babel

Full text

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

Local equations for the generalized Lotka-Volterra model on sparse asymmetric graphs Preprint: APS/123-QED D. Machado, P. Valigi, T. Tonolo, M. C. Angelini Abstract Real ecosystems are characterized by sparse and asymmetric interactions, posing a major challenge to theoretical analysis. We introduce a new method to study the generalized Lotka-Volterra model with stochastic dynamics on sparse graphs. By deriving local Fokker-Planck equations and employing a mean-field closure, we can efficiently compute stationary states for both symmetric and asymmetric interactions. We validate our approach by comparing the results with the direct integration of the dynamical equations and by reproducing known results and, for the first time, we map the phase diagram for sparse asymmetric networks. Our framework provides a versatile tool for exploring stability in realistic ecological communities and can be generalized to applications in different contexts, such as economics and evolutionary game theory. I Introduction The stability of complex ecosystems and the rules governing species coexistence present a central puzzle in theoretical ecology. The generalized Lotka-Volterra (gLV) model has been a cornerstone of this inquiry, also related to models used in evolutionary game theory and in economic theory 43 ; 20 ; 24 . Recently, the random, symmetric, and dense version of the gLV model has been analyzed through the lens of equilibrium statistical mechanics. This approach has yielded profound insights, revealing how the phases of single equilibrium, unbounded growth, and multiple equilibria are dictated by a handful of macroscopic parameters: the mean \mu and variance \sigma of inter-species interactions, and the intensity of demographic noise T . A key finding is that densely connected networks with random symmetric interactions can exhibit a phase of multiple equilibria at high interaction heterogeneity 3 . However, real ecological networks are generally neither dense nor symmetric. In real ecosystems, the interactions between two species are almost always asymmetric, the existence of predator-prey couples of species being just an example. For this reason, some other works have tried to go beyond the assumption of symmetry 13 ; 21 ; 50 ; 51 ; 48 ; 10 ; 27 ; 7 ; 25 ; 47 . Moreover, in real ecosystems, a species typically interacts only with a few others 19 ; 14 . In the last years, Random Matrix Theory (RMT) results 37 ; 56 have shown that the spectra of sparse random graphs exhibit qualitative differences from the dense case, suggesting that this may have implications for the stability of ecological models defined on sparse networks. Accordingly, it would be interesting to study the properties of the gLV model on graphs with finite connectivity. Unfortunately, in this case, the methods used for fully-connected systems 3 ; 50 cannot be applied, in particular because no central-limit-type arguments hold. In a very recent paper, for the first time, the equilibrium properties of the symmetric gLV model on a sparse graph were analyzed 54 using the so-called Belief-Propagation (BP) cavity method. This can, however, be used only when an equilibrium measure exists, preventing its implementation in the case of asymmetric interactions. In general, the description of out-of-equilibrium systems has been even more elusive, and the available techniques deal only with specific limits. When species interact through a fully-connected network 52 , we can use Dynamical Mean-Field Theory (DMFT) to describe the temporal evolution of dynamic observables, which has also been extended to non-Gaussian disorder in the interactions 6 . Other generalizations include the limit of very small connectivity 6 , the limit of a large connectivity that grows sublinearly with the number of species in the graph 1 , and the case of unidirectional interactions 41 . However, until now, there has been no way to systematically analyze gLV models with sparse asymmetric interactions. In this work, we bridge this gap by analyzing both symmetric and asymmetric, quenched-disordered sparse interactions. We go beyond equilibrium statistical mechanics, introducing a new method for the evaluation of the stationary probability distribution for the stochastic differential equations (SDE) that describe the evolution of the species abundance. We start from the usual formulation of the stochastic dynamics in terms of an SDE and derive the equivalent Fokker-Planck equation for the associated time-dependent probability densities. However, solving the full system of partial differential equations defined on the whole graph in a high-dimensional space is a cumbersome task. To overcome this difficulty, we derive local closures that allow us to obtain tractable relations for the stationary distributions. The main idea of a local closure is to propose an ansatz for the probability densities, in general involving some suitable factorizations that exploit the properties of the interaction graph. It has been used successfully in several contexts, such as the study of epidemics spreading on networks 57 ; 16 ; 46 ; 44 , algorithmic dynamics in hard combinatorial optimization problems 5 ; 34 , spin-glass dynamics in random graphs 35 , or the dynamics of the voter model 29 . As far as we know, this work constitutes the first application of local closures to Fokker-Planck equations in sparse graphs. The new approximate descriptions derived here are what we call local Fokker-Planck equations. We will validate our method by applying it to different situations, comparing its prediction with the results obtained from the direct numerical integration of the SDE and recovering known results from previous literature. We also show how, starting from the general equations in the asymmetric case, one can recover the BP equations when only symmetric couplings are considered. The rest of the manuscript is organized as follows. In Section II we introduce the generalized Lotka-Volterra equations and the underlying networks that we will analyze. In Section III , we present the Fokker-Planck equations that describe the evolution of the probability distributions of the species abundances in time. These are complicated global equations, which in general are not solvable. For this reason, in Section III.1 we introduce a new local closure, corresponding to a mean-field approximation for the dynamics, called Individual Based Mean Field (IBMF). Our procedure leads to the main local solvable Fokker-Planck equations that we discuss in this article. In Section III.2 we go beyond IBMF, introducing a more refined closed local Fokker-Planck equation that we call Pair Based Mean Field (PBMF), showing that the BP equations introduced in Ref. 54 correspond to the stationary solution of PBMF in the symmetric case. We validate our new methods in Section IV , where we give details on the numerical implementation of IBMF both at null and finite temperatures. We compare the stationary abundances obtained with IBMF with those obtained from simulations in the presence of thermal noise for a single random graph. The results are a useful example to emphasize the strengths of IBMF, and to also point out its limitations. Then, we proceed to apply IBMF to three different scenarios. In Subsection IV.1 , we study undirected graphs with asymmetric interactions at null temperature, obtaining the corresponding phase diagram in the plane (\mu,\sigma) for the first time, as far as we know. We thus generalize the results for the fully-connected asymmetric case in Ref. 13 and for the sparse case with symmetric interactions in Ref. 54 . In Subsection IV.2 , we apply the IBMF closure to directed graphs with null variance

Read the complete text