Thesis
Evolution PDEs on graphs and inverse problems in aggregation-diffusion equations
- Abstract:
-
This thesis contains contributions in two areas of research and is therefore divided into two parts. On the one hand we study evolution partial differential equations (PDEs) on graphs, and on the other, we address the inverse problem of learning interaction potentials in aggregation-diffusion equations.
In Part One, we investigate evolution PDEs on co-evolving graphs, i.e. a setting in which the topology of the graph evolves in time with its dynamics coupled to the dynamics on the graph. We establish the well-posedness of the system and investigate scenarios where the graph evolves at different rates—slower or faster—than the dynamics on the graph. Additionally, we derive a discrete-to-continuum limit as the number of vertices tends to infinity. We lift the equation on the co-evolving graph to a type of nonlinear continuity equation on Euclidean space whose velocity field depends on the graph structure. We obtain its well-posedness and comment on discrete-to-continuum limits as well as on the contractivity of solutions of this equation. Furthermore, we describe the long-time behaviour of equations on co-evolving graphs with a particular type of velocity leading to diffusive dynamics.
Part Two addresses the inverse problem of learning the interaction potential of an aggregation-diffusion equation. We use the PDE's numerical solutions as data and propose a quadratic error functional enhanced with sparsity promoting regularization to solve the learning problem. We obtain stability estimates showing that this error functional controls the 2-Wasserstein distance between solutions generated by the true and learned interaction potential and show the effectiveness of our approach with numerical examples in one and two dimensions.
Actions
Access Document
- Files:
-
-
(Preview, Dissemination version, pdf, 2.5MB, Terms of use)
-
Authors
Contributors
- Institution:
- University of Oxford
- Division:
- MPLS
- Department:
- Mathematical Institute
- Role:
- Supervisor
- Institution:
- University of L'Aquila
- Role:
- Supervisor
- Funding agency for:
- Esposito, A
- Funding agency for:
- Mikolás, L
- Funding agency for:
- Carrillo, JA
- Grant:
- 883363
- Programme:
- Advanced Grant Nonlocal-CPD (Nonlocal PDEs for Complex Particle Dynamics: Phase Transitions, Patterns and Synchronization)
- Funding agency for:
- Mikolás, L
- Grant:
- EP/S023925/1
- Programme:
- EPSRC Centre for Doctoral Training in Mathematics of Random Systems: Analysis, Modelling and Simulation
- DOI:
- Type of award:
- DPhil
- Level of award:
- Doctoral
- Awarding institution:
- University of Oxford
- Language:
-
English
- Keywords:
- Subjects:
- Deposit date:
-
2025-02-09
- ARK identifier:
Terms of use
- Copyright holder:
- László Mikolás
- Copyright date:
- 2024
If you are the owner of this record, you can report an update to it here: Report update to this record