Descrizione del progetto
Un sistema scalabile e sicuro per l’intelligenza artificiale con accelerazione hardware
I dati vengono raccolti ovunque tutto intorno a noi, registrando volumi in rapida crescita provenienti da innumerevoli fonti. Il problema è come estrapolare conoscenze preziose e valore commerciale dai dati. Questa operazione richiede metodi, approcci e paradigmi di ingegnerizzazione nuovi negli ambiti dell’apprendimento automatico, dell’analisi e della gestione dei dati. Il progetto EVEREST, finanziato dall’UE, sta elaborando un approccio olistico per il calcolo e la comunicazione della co-progettazione all’interno di un sistema eterogeneo, distribuito, scalabile e sicuro per analisi di megadati ad alte prestazioni. Ciò semplificherà la programmabilità di architetture eterogenee e distribuite tramite un approccio di progettazione «basato sui dati» nonché avvalendosi dell’intelligenza artificiale con accelerazione hardware e di un paradigma hardware/software unificato. Il progetto convaliderà il suo approccio applicandolo in scenari aziendali reali, tra cui un modello di previsione meteo basato sull’analisi e una struttura intelligente di modellizzazione del traffico cittadino.
Obiettivo
The distributed and heterogeneous nature of the data sources in High Performance Big Data Analytics (HPDA) applications, as well as the required computational power, is pushing designers towards novel computing systems that combine HPC, Cloud, and IoT solutions (for efficient and distributed computation closer to the data) with Artificial Intelligence (AI) algorithms (for knowledge extraction and decision making).
In this context, the EVEREST project addresses the matching problem between application (and data) requirements, and the characteristics of the underlying heterogeneous hardware. Only an optimal match leads to efficient computation. In particular, we forecast that the creation of future Big Data systems will be of course data-driven, but also featuring complex heterogeneous and reconfigurable architectures that must be redesigned or customized based on the nature and locality of the data, and the type of learning/decisions to be performed.
The EVEREST project aims at developing a holistic approach for co-designing computation and communication in a heterogeneous, distributed, scalable and secure system for HPDA. This is achieved by simplifying the programmability of heterogeneous and distributed architectures through a “data-driven” design approach, the use of hardware-accelerated AI, and through an efficient monitoring of the execution with a unified hardware/software paradigm. EVEREST proposes a design environment that combines state-of-the-art, stable programming models, and emerging communication standards, with novel and dedicated domain-specific extensions.
Three industry-relevant application scenarios are used to validate the EVEREST approach and act as business cases for the project exploitation: (i) a weather analysis-based prediction model for the renewable energy trading market, (ii) an application for air-quality monitoring of industrial sites, and (iii) a real-time traffic modeling framework for intelligent transportation in smart cities.
Campo scientifico
- natural sciencescomputer and information sciencesartificial intelligence
- engineering and technologycivil engineeringurban engineeringsmart cities
- natural sciencescomputer and information sciencesinternetinternet of things
- engineering and technologyenvironmental engineeringenergy and fuelsrenewable energy
- natural sciencescomputer and information sciencesdata sciencebig data
Parole chiave
Programma(i)
Meccanismo di finanziamento
RIA - Research and Innovation actionCoordinatore
8803 Rueschlikon
Svizzera