Thibault Lahire (Dassault Aviation) presents Deep Learning research at ICLR 2025

Thibault Lahire of Dassault Aviation, an EICACS project partner, recently presented his latest research at the International Conference on Learning Representations (ICLR 2025) held in Singapore.

Thibault Lahire (Dassault Aviation)

The paper, titled "Towards a Better Use of Importance Sampling for Deep Neural Networks Training," explores how smarter data sampling strategies can enhance the training efficiency of deep neural networks.

The study introduces a new metric to assess the effectiveness of sampling schemes and examines how these interact with optimization algorithms, offering practical insights to improve deep learning performance. These findings are highly relevant to mission-critical AI applications, where performance, efficiency, and reliability are essential.

Dassault Aviation’s participation at ICLR reflects EICACS’s broader commitment to integrating high-level AI research into the evolution of next-generation European defense technologies.

Stay tuned for more updates from EICACS partners at the intersection of science, technology, and strategic innovation.