Can we trust the prediction of a neural network? 🤔
This was the question addressed by [Esa Ollila], a professor at Aalto University (Finland), during his seminar at TéSA this Tuesday, October 20. The laboratory has been pleased to host him for several weeks 🎓
Holder of a PhD in statistics and signal processing, he is the deputy head of the Department of Information and Communications Engineering at Aalto University. He also sits on the board of directors of EURASIP (European Association for Signal Processing).
His answer: conformal prediction 📐, a method that:
• adapts to any artificial intelligence model without modifying it
• associates a set of possible outcomes with each prediction, with a reliability guarantee
• remains valid even with a limited volume of data
For systems where every decision matters, such as space 🛰️ or aerospace ✈️, having a reliable measure of uncertainty is a major asset.
A specialist in robust statistics, @Esa Ollila is a co-author of the book “Robust Statistics for Signal Processing” (Cambridge University Press, 2018) 📘. His presentation, illustrated by a Python notebook 💻, concluded with applications in signal processing.
Thank you for this contribution!
Are you interested in our research work 🔬?
Find us at https://www.tesa.prd.fr 🛰️, we would be delighted to exchange with you!
#TeSA
#Doctorat
#Recherche
#TraitementDuSignal