Ghouthi Boukli Hacene was awarded the prize for the best thesis of the Futur & Ruptures 2020 (future and disruptive innovation) Mines-Télécom Fondation program (in joint first place with Laetitia Jeancolas of Télécom SudParis) for his work on the "Processing and learning of deep neural networks on chip" conducted within the Electronics Department at IMT Atlantique and within the framework of the PRACOM chaire (French research consortium).
PRACOM is a reference centre for the development of communications and related systems. Its original model is based on the membership of its members (for 3 renewable years) and collegial management (4 heads of departments in rotation: Signal and Communications, Electronics, Microwaves and SRCD (Network Systems, Cybersecurity and Digital Law)). Over time, it has built up lasting collaborations between IMT Atlantique and industry thanks to several partners. The current members of PRACOM are TurboConcept, Merce (Mitsubishi), SAFRAN Data Systems, Engie-Ineo, ANRT-Maroc, Interface Concept, and MVG (Satimo). The research consortium also benefits from the support of local authorities and the European Union.
1st prize thesis by Ghouthi Boukli Hacene (Electronics department)
The subject of Ghouthi Boukli Hacene's thesis fits perfectly with PRACOM's missions since it involves not only fundamental research questions on machine learning, but also addresses issues hampering industrial applications especially in an embedded context. Indeed "in the field of machine learning, if deep neural networks have become standard in almost all fields, we know that to achieve the best performance they rely on a number of important parameters involving high complexity in terms of memory and computation. In my thesis, I therefore sought to compress neural networks and reduce this complexity in order to make their use in mobile applications possible," explains Ghouthi, who proposed several algorithms that are less energy-hungry. They are also more respectful of data privacy as they can be embedded (in smartphones, connected objects, or even cars).
Ghouti Boukli Hacene presents his work
"Ghouthi has managed to propose very promising original results which have been praised by (AI expert and thesis jury member) Joshua Bengio, from MILA" underlines his thesis supervisor, Michel Jézéquel, head of the Electronics Department. His skills in the field of machine learning have also earned him the 1st prize in the AI thesis 2020 from the French Association for Artificial Intelligence (AFIA) and his participation in the Artificial Intelligence Platform from June 29 to July 3, 2020 in Angers PFIA 2020".
PRACOM members are delighted to receive these awards, which fund around 2 theses per year and aim to develop more collaborative actions to bring new research projects to the fore. Ghouthi Boukli Hacene is currently pursuing a post-doc as he wishes to "continue in scientific research to contribute to the development of AI and its integration into everyday life."
Relive the award ceremony (in 100% digital format due to the COVID-19 health crisis) on the Mines Telecom Foundation's YouTube channel: Watch the video
by Pierre-Hervé VAILLANT