A team of researchers from ICAE participates in four out of seven working packages (WP) of the ENFIELD (HORIZON-CL4-2022-HUMAN-02) project.
ENFIELD project is set to contribute to the establishment of a unique European Centre of Excellence dedicated to advancing fundamental research in Adaptive, Green, Human-Centric, and Trustworthy AI. This initiative, crucial for successful AI development and deployment in Europe, aims to attract top talents, technologies, and resources from renowned research and industry players across the continent. By addressing industry challenges in healthcare, energy, manufacturing, and space, ENFIELD endeavors to strengthen the EU's competitive position in AI while generating significant socio-economic impact. With 30 consortium members from 18 countries, including leading research organizations, businesses, SMEs, and public sector representatives, ENFIELD fosters collaboration on critical research and innovation frontiers in AI. Through the creation of unique AI solutions, high-impact publications, strategic documents, and innovative exchange schemes, researchers from ICAE and other institutions contribute to shaping the future of AI in Europe.
UCM Reserach Team contributions to the Green AI Pillar (WP2: Research Exellence)
Dounload GUI.exe v2.0.0 from UCM cloud
Eco-RETINA Graphical User Interface (GUI.exe) v2.0.0 at GitHub
EcoRETINA Operational Tutorial v2.0.0
Eco-RETINA Graphical User Interface (GUI) v1.0.0 at GitHub
Functioning as a regression-based flexible approximator, it is linear in parameters but nonlinear in inputs, employing a selective model search to optimize performance. The algorithm manages multicollinearity while emphasizing speed, accuracy, and environmental sustainability. Its modular and transparent structure facilitates easy interpretation and modification, making it an invaluable tool for researchers in developing explicit models for out-of-sample forecasting. The algorithm generates outputs such as a list of relevant transformed inputs, coefficients, standard deviations, and confidence intervals, enhancing its interpretability.
Now implemented in Python and soon available on GitHub, Eco-Retina introduces several new features, including measuring CO₂ emissions and energy consumption. These enhancements, alongside improved data transformations, bottleneck elimination, and a user-friendly interface, significantly boost its performance. The algorithm achieves remarkable reductions in carbon footprint and power consumption (ranging from 50% to 90%) while significantly reducing computational time. Empirical results indicate that Eco-Retina is not only a sustainable alternative to conventional neural networks but also surpasses them in certain aspects, offering a competitive edge in accuracy, interpretability, and environmental impact.
The paper has been submitted to two congresses (waiting for acceptance) and will be submitted to a journal soon. We are about to finish the complete deployment on GitHub and the GUI (Graphical User Interface).
Consortium Partners
- 18 countries: Austria, Cyprus, Denmark, Estonia, Finland, France (3), Germany, Greece, Hungary, Italy (3), Netherlands (2), Norway (4), Portugal (4), Romania, SIovenia, Spain (2), Sweden, United Kingdom
- 30 partners
- 36 months
- 4 third party open calls
- € 11.487.793,7 Funding
Research Team
|
Associate Professor
|
|
|
Doble Grado en Economía - Matemáticas y Estadística (UCM) E-mail: |
|
|
Full Professor
|
|
|
CAPILLA ROMEROSA, MANUEL JAVIER Doble Grado en Economía - Matemáticas y Estadística (UCM) E-mail: |
|
|
Assistant Professor E-mail: Contact: 91 394 24 74 / 212A - pabellón de 2º curso Field/s: Digital Economics and Applied Econometrics.
|
|
|
ELEK MAXIME Mines Saint-Étienne E-mail: |
|
|
BERNADAC jules Mines Saint-Étienne E-mail: |
|
|
|
