Skip to content

We Collaborate on the TAUDoS Project

Before being a trendy expression, “Trustful AI” represents a primary need in machine learning: how can I trust a model and its decisions? Without an answer to such a question, the adoption of machine learning models is complicated.

To address this problem, we started a new collaboration with Aix-Marseille University, Montreal University, Nantes University, and St-Etienne on a four-year project called TAUDoS. Our research engineer Nicolas explains:

“Amongst other things, trust can be brought by interpretability: “What made my model make a certain decision?Explaining decisions is especially hard for models trained on sequential data (time series, natural language, …), and there is still a lot of work to do in order to have a general well-behaved solution. Finding such an approach is the aim of the Taudos project.”
Last week, the whole consortium met in Marseille, to discuss findings, future directions, and challenge ideas. Multiple talks for both theoretical and industrial settings have been presented. The main research direction is to find links between automatas (well-known interpretable models created a long time ago) and modern, less interpretable approaches.”

 

 
The TAUDoS project also supports the ambition of our two-year research program, BISHOP, which aims at addressing the challenges of responsible artificial intelligence, and ensure data confidentiality and trust in AI models

Releated Posts

Augment to Interpret: Unsupervised and Inherently Interpretable Graph Embeddings

In this paper, we study graph representation learning and show that data augmentation that preserves semantics can be learned and used to produce interpretations. Our framework, which we named INGENIOUS, creates inherently interpretable embeddings and eliminates the need for costly additional post-hoc analysis.
Read More

SANGEA: Scalable and Attributed Network Generation

In this paper, we present SANGEA, a sizeable synthetic graph generation framework that extends the applicability of any SGG to large graphs. By first splitting the large graph into communities, SANGEA trains one SGG per community, then links the community graphs back together to create a synthetic large graph.
Read More