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
This report will summarize the main results of Task 4.1, which is a review of state-of-the-art solutions for implementing software interface layers for ML applications in space. Interface layers are needed to connect high-level frameworks such as TensorFlow, PyTorch or ONNX with the testbed hardware. The outcome of this review will be used to identify the most efficient solution for the interface layer, which will be designed and developed in Task 4.2.
Report on the performances of the ML methods selected to be ported on the testbed (si apre in una nuova finestra)Within WP 2, algorithms for the enhancement of the capabilities for on-board science operation and applications of space mission are developed on consumer-grade computing systems. These algorithms will regard: autonomous triggering of special measurement modes; the selective downlink of plasma environment parameters; on-board data analysis of three-dimensional particle distribution function; on-board prediction of SEP; and image analysis algorithms. In the report, algorithms and algorithms performance results will described.
Project website and dissemination and communication plan. (si apre in una nuova finestra)This deliverable reports on the setting up and organization of the website and provides the first dissemination and communication plan to be updated during the project.
Testbed Interface Layer report (si apre in una nuova finestra)This report will include the design and development of a software layer aiming to port the high-level models (e.g. TF, PyTorch, ONNX) to the soft-GPU. The soft-GPU is an application-independent FPGA design that speeds up the inference of the models ported via the virtual layer. The soft-GPU will be developed featuring HDL language, providing the possibility to be implemented in different radiation-tolerant FPGAs used for space applications.
Hardware Feasability Study (si apre in una nuova finestra)This deliverable is the outcome of the feasibility study and analysis of AI/ML methods (of WP2) and their materialization on FPGAs. Here we will look at the different AI/ML methods used in WP2 and understand which of these should be accelerated using the FPGAs. We will apply performance-analysis workflows ( e.g., roofline analysis, intensity, etc.) to understand which parts should be accelerated in order to maximize the performance and use of the acceleration.The outcome is a public report.
Report specifying the requirements for the design of the testbed (si apre in una nuova finestra)Different possible in-flight scenarios have to be taken into consideration where efficient functioning of the algorithms is desirable. Therefore, a requirements analysis will be performed that will serve for an effective testbed design both in terms of hardware and the software. The functional (software and hardware features) and non-functional (performance) requirements will be described in the report.
Survey of state-of-the-art hardware methods (si apre in una nuova finestra)This deliverable is the outcome of the survey of state-of-the-art methods for exploiting reconfigurable architectures to accelerator AI/ML workloads with a particular focus on doing so in space. More specifically, we will look at understanding the unique requirements that operation in space demands in order to provide a remedy for them in the remaining parts of the work. In particular, we will be looking at what kind of AI/ML implementation is most suitable (e.g., spiking- or rate-based), the type of resilience that may be needed (e.g., redundancy), as well as different number representations.The outcome of this deliverable is a public report.
Yearly report on dissemination, communication, and exploitation plans. (si apre in una nuova finestra)This deliverable is an update of D6.1.
Pubblicazioni
Autori:
Mario Imbrogno, Claudio Meringolo, Alejandro Cruz-Osorio, Luciano Rezzolla, Benoît Cerutti, Sergio Servidio
Pubblicato in:
The Astrophysical Journal Letters, Numero 990, 2025, ISSN 2041-8205
Editore:
American Astronomical Society
DOI:
10.3847/2041-8213/ADFB4C
Autori:
Beniamino Sanò, Nathan N. Maes Anno, David L. Newman, Marty Goldman, Francesco Valentini, Denise Perrone, Giovanni Lapenta
Pubblicato in:
Journal of Geophysical Research: Machine Learning and Computation, Numero 2, 2025, ISSN 2993-5210
Editore:
American Geophysical Union (AGU)
DOI:
10.1029/2024JH000233
Autori:
D Trotta, F Valentini, D Burgess, S Servidio
Pubblicato in:
Monthly Notices of the Royal Astronomical Society, Numero 536, 2024, ISSN 0035-8711
Editore:
Oxford University Press (OUP)
DOI:
10.1093/mnras/stae2750
Autori:
F. Valentini, T. M. O'Neil, D. H. Dubin
Pubblicato in:
Physics of Plasmas, Numero 32, 2025, ISSN 1070-664X
Editore:
AIP Publishing
DOI:
10.1063/5.0256797
Autori:
S. Zanelli; S. Perri; M. Condoluci; P. Veltri; F. Pegoraro; O. Pezzi; D. Perrone; D. Trotta; F. Valentini
Pubblicato in:
Physics of Plasmas, Numero 32, 2025, ISSN 1089-7674
Editore:
America Institute of Physics
DOI:
10.1063/5.0259317
Autori:
Jonah Ekelund, Ricardo Vinuesa, Yuri Khotyaintsev, Pierre Henri, Gian Luca Delzanno, Stefano Markidis
Pubblicato in:
2024 IEEE 20th International Conference on e-Science (e-Science), 2024
Editore:
IEEE
DOI:
10.1109/e-Science62913.2024.10678688
È in corso la ricerca di dati su OpenAIRE...
Si è verificato un errore durante la ricerca dei dati su OpenAIRE
Nessun risultato disponibile