The formal verification of neural networks presents many challenges. Although there are languages to describe how a neural network is expected to behave, current validation tools do not consider the full richness of these languages. CEA-List is investigating practices from the field of programming languages to expand the scope of what can be formally verified.
Read more
We developed a method, PEM, to adapt the Monte Carlo algorithm and penalize regions that are uncertain for the AI, mitigating AI errors and making deep-learning-based materials modeling trustworthy.
Read more
HPC and AI applications are increasingly data intensive, their memory access patterns are irregular, making the cache memory less efficient. HPDcache, which allows multiple memory requests to be executed simultaneously, solves this problem.
Read more
Kentyou’s mission is twofold: firstly, to facilitate data access, unification, and processing. Secondly, to help communities obtain actionable information for decision-making.
Read more
Developed from fifteen years of R&D at the CEA, ASTERIOS® software tools are protected by several patents. Initially industrialized by the start-up Krono-Safe, it was acquired in 2023 by Safran Electronics & Defense, its first client.
Read more
AI models are rapidly making inroads into a wide range of industrial use cases. CEA-List has developed automated testing software called AIMOS that leverages formal methods to make reliability testing simpler and easier. The software was developed under the multi-partner Confiance.ai trustworthy AI program, part of the “France 2030” national investment plan.
Read more