Freddy Lecue

Uncovering the Semantics of Deep Neural Networks with Knowledge Graphs

We are currently seeing is a surge of innovation and uptake focused on machine learning, and more specifically deep learning — which is most successful in low-level pattern recognition tasks from many digitalized content such as image, video, speech or text. Today’s machine learning systems are achieving impressive results, having demonstrated wide applicability with real-world impact in many contexts. Latest results are GitHub Copilot powered by Open AI Codex and its customized version of GPT-3, or Gato from DeepMind, which works as a multi-modal, multi-task, multi-embodiment generalist policy, able to generalize beyond expectation. Powered by deep neural networks, the inner semantics of their mechanics remains largely opaque an open to lots of unanswered questions, making the engineering of any new architecture very brittle. This presentation will present some results and directions towards uncovering the semantics of deep neural networks using knowledge graphs to scale the engineering of deep neural networks. 

back to overview

Watch Recording


Dr. Freddy Lecue is the Chief Artificial Intelligence (AI) Scientist at CortAIx (Centre of Research & Technology in Artificial Intelligence eXpertise) at Thales in Montreal - Canada. He is also a research associate at INRIA, in WIMMICS, Sophia Antipolis - France. Before joining the new R&T lab of Thales dedicated to AI, he was AI R&D Lead at Accenture Labs in Ireland from 2016 to 2018. Prior joining Accenture he was a research scientist, lead investigator in large scale reasoning systems at IBM Research from 2011 to 2016, a research fellow at The University of Manchester from 2008 to 2011 and research engineer at Orange Labs from 2005 to 2008. His research area is at the frontier of intelligent / learning / reasoning systems, and Internet of Things. He has a strong interest on Explainable AI i.e., AI systems, models and results which can be explained to human and business experts, as well as systems which combine learning and reasoning capabilities.