Skip to main content

Research Repository

Advanced Search

Noise reduction in coarse bifurcation analysis of stochastic agent-based models: an example of consumer lock-in

Avitabile, Daniele; Hoyle, Rebecca; Samaey, Giovanni

Authors

Daniele Avitabile

Rebecca Hoyle

Giovanni Samaey



Abstract

We investigate the occurrence of coarse macroscopic states in an agent-based model of consumer lock-in. The system studied here is a modification of an existing model by Garlic and Chli [24] and it serves as a prototypical Ising-type sociological system with binary state variables and
spatially-dependent agent parameters. In the regime of globally-coupled agents with independent identically-distributed parameters, we derive an analytic approximate coarse evolution-map for the expectation of the average purchase. Following Barkley et al. [5], we interpret metastable locked-in states as fixed points of this one-dimensional first moment map. We then study the emergence of coarse fronts in the regime of heterogeneous agents with strongly discordant preferences. When agent polarization becomes less pronounced, the front destabilizes and one of the two products prevails, giving rise to inhomogeneous profiles featuring pockets of resistance. Stochastic continuation of the spatially-extended case poses a numerical challenge, as Jacobian-vector products are severely affected by noise. We exploit the non-uniqueness of the lifting step introducing weighted lifting/restriction operators, which result in variance-reduced Jacobian-vector products. We test our numerical strategy and show that weighted operators induce good convergence properties of the Newton-GMRES solver. We then show that macroscopic fronts destabilise at a coarse
symmetry-breaking bifurcation.

Citation

Avitabile, D., Hoyle, R., & Samaey, G. (2014). Noise reduction in coarse bifurcation analysis of stochastic agent-based models: an example of consumer lock-in. SIAM Journal on Applied Dynamical Systems, 13(4), 1583-1619. https://doi.org/10.1137/140962188

Journal Article Type Article
Acceptance Date Sep 3, 2014
Publication Date Nov 18, 2014
Deposit Date Mar 28, 2014
Publicly Available Date Nov 18, 2014
Journal SIAM Journal on Applied Dynamical System
Print ISSN 1536-0040
Electronic ISSN 1536-0040
Publisher Society for Industrial and Applied Mathematics
Peer Reviewed Not Peer Reviewed
Volume 13
Issue 4
Pages 1583-1619
DOI https://doi.org/10.1137/140962188
Public URL https://nottingham-repository.worktribe.com/output/997924
Publisher URL https://epubs.siam.org/doi/10.1137/140962188
Additional Information Copyright: Society for Industrial and Applied Mathematics

Files





Downloadable Citations