Real-time inversion using self-explainable deep learning driven by expert knowledge

Informazioni relative al progetto

IN-DEEP

ID dell’accordo di sovvenzione: 101119556

DOI

10.3030/101119556

Data della firma CE 13 Giugno 2023

Data di avvio 1 Febbraio 2024

Data di completamento 31 Gennaio 2028

Finanziato da

• Marie Skłodowska-Curie Actions (MSCA)

Costo totale

Nessun dato

Contributo UE

€ 2 046 614,40

Coordinato da

UNIVERSIDAD DEL PAIS VASCO/ EUSKAL HERRIKO UNIBERTSITATEA
Spain

CORDIS fornisce collegamenti ai risultati finali pubblici e alle pubblicazioni dei progetti ORIZZONTE.

I link ai risultati e alle pubblicazioni dei progetti del 7° PQ, così come i link ad alcuni tipi di risultati specifici come dataset e software, sono recuperati dinamicamente da .OpenAIRE .

Risultati finali

Documents, reports (9)

Theoretical advances for Deep PDE solvers.

Initial report about the main theoretical results and comparison between the different Deep PDE solvers.

Baseline model for PDE enhancement with collected data.

Characterization of the techniques developed to enrich PDEs utilizing available data.

Datasets requirements for parametric PDEs.

Generation, analysis, and sampling of the dataset in parametric PDEs.

Gender and diversity equality plan.

Priorities, concrete objectives, and the specific measures that will be implemented to improve gender and diversity equality within the IN-DEEP project.

Training events in M1-24

Description and self-assessment of training events and proposed correction measures.

Computational complexity control through adaptive strategies in Deep PDE solvers for parametric problems.

Efficiency assessment of the novel adaptive strategies developed for Deep PDE solvers applied to parametric problems.

Optimization techniques for Deep PDE solvers.

Analysis of the effect of adaptive strategies on the algorithms.

Advances in theory and numerical approximation of Deep PDE solvers for parametric problems.

Initial Analytical study and partial numerical validation of the proposed methodology. Comparison with conventional solvers.

Loss function choice criteria, according to the PDE problem.

Study of the appropriateness and efficiency of the loss function in Deep PDE solvers. Advantages, limitations, recommendations for forward, inverse, and parametric problems.

Websites, patent fillings, videos etc. (1)

Website

Project website, logo and social media page creation

Pubblicazioni

Book chapters (2)

EXPBrain: Exponential Integrators for Glioblastoma Brain Tumor Simulations

Autori: Magdalena Pabisz, Dominika Ciupek, Askold Vilkha, Maciej Paszyński
Pubblicato in: Lecture Notes in Computer Science, Computational Science – ICCS 2025 Workshops, 2025, ISBN 978-3-031-97554-7
Editore: Springer Nature Switzerland
DOI: 10.1007/978-3-031-97554-7_10

Tensorial Implementation for Robust Variational Physics-Informed Neural Networks

Autori: Askold Vilkha, Carlos Uriarte, Paweł Maczuga, Tomasz Służalec, Maciej Paszyński
Pubblicato in: Lecture Notes in Computer Science, Computational Science – ICCS 2025, 2025, ISBN 978-3-031-97626-1
Editore: Springer Nature Switzerland
DOI: 10.1007/978-3-031-97626-1_5

Peer reviewed articles (2)

Collocation-based robust variational physics-informed neural networks (CRVPINNs)

Autori: Marcin Łoś, Tomasz Służalec, Paweł Maczuga, Askold Vilkha, Carlos Uriarte, Maciej Paszyński
Pubblicato in: Computers & Structures, Numero 316, 2025, ISSN 0045-7949
Editore: Elsevier BV
DOI: 10.1016/J.COMPSTRUC.2025.107839

Computer Methods in Applied Mechanics and Engineering

Autori: Shima Baharlouei; Jamie M. Taylor; Carlos Uriarte; David Pardo
Pubblicato in: Computer Methods in Applied Mechanics and Engineering, 2024, ISSN 0045-7825
Editore: Elsevier BV
DOI: 10.2139/SSRN.5004047