IRT Saint Exupéry • Technological Research Institute

Accelerating science, technology research & transfers to industry

DEEL: The Robust and Explainable Artificial Intelligence Program Enters Phase 2

DEEL (DEpendable and Explainable Learning) is a project, primarily a collaborative Franco-Quebec research team based in Toulouse (France) and Montreal (Quebec). Launched in 2018, its main objective is to develop new solutions for reliable and explainable machine learning.

DEEL is managed by IVADO, CRIAQ, IID, ANITI, and IRT Saint Exupéry. With a budget of €30 million, 27 partners including 15 from the industry, the first phase of the DEEL project (2018-2023) produced over 80 publications (30% in A or A* conferences or journals), 8 autonomous libraries (6 open source), and over 100 deliveries to industrial partners. DEEL organizes an annual conference named Mobilit’AI, a monthly online seminar, several summer schools, and contributes to the AIBT master’s program co-organized with SUPAERO.

The DEEL project is inspired by the research model of major digital technology companies: publications, open source, and rapid transfer from research to operational applications. It brings together researchers, data scientists, engineers, and experts from several industrial partners at the same location to address the topic of machine learning for critical systems. This represents an ecosystem of over 150 people today. Researchers’ work is protected while encouraging them to focus on “scientific challenges” inspired by the industry. Teamwork and visibility are sought after, and results are continuously transferred to industrial partners.

Ana Gonzalez – project manager DEEL, Lionel Cordesses – Director of Intelligent Technologies, and Grégory Flandin – Programme Director

The aim of the project is to provide industrial partners with tools and technological components based on artificial intelligence (AI).

These tools will enable manufacturers to secure the development, within a short timeframe, of their critical systems incorporating AI functions. Several sub-objectives have been defined to achieve this:

  • the first objective of the project is to create visible and identifiable global scientific and technological knowledge on the subjects of AI that is reliable and explainable, in a context of statistical machine learning;
  • The second objective is to create the conditions within the project itself for a rapid transfer of this work to industrial partners;
  • the third objective is to publicize the scientific challenges developed with the industrial partners, in order to develop the interest of the international community and attract relevant work.

Since 2018, DEEL has been a cornerstone project that has fostered closer ties between local authorities of Toulouse Métropole and the city of Montreal, supported by the Fonds Franco-Québécois pour la Coopération Décentralisée (FFQCD).

This collaboration has energized innovation in the field of critical AI, by valuing the convergence between the mobility (aerospace, automotive, rail) and digital ecosystems of the two territories.

Buoyed by its successes in phase 1 from 2018 to 2023, and with renewed support from its industrial and academic partners, phase 2 of DEEL begins in January 2024 for an additional four years. It continues its goal of certifying AI-based systems across several axes: explainability of AIs, data biases, theoretical guarantees and uncertainty quantification, robustness of algorithms, reinforcement learning, embeddability. In particular, it will extend the results of the first phase to textual and multimodal data (e.g., images and text).

DEEL: The Robust and Explainable Artificial Intelligence Program Enters Phase 2
Scroll to top