Numerical Uncertainty of Convolutional Neural Networks Inference for Structural Brain MRI Analysis

no code implementations • 3 Aug 2023 • Inés Gonzalez Pepe, Vinuyan Sivakolunthu, Hae Lang Park, Yohan Chatelain, Tristan Glatard

This paper investigates the numerical uncertainty of Convolutional Neural Networks (CNNs) inference for structural brain MRI analysis.

Brain Segmentation Segmentation

Numerical Stability of DeepGOPlus Inference

1 code implementation • 13 Dec 2022 • Inés Gonzalez Pepe, Yohan Chatelain, Gregory Kiar, Tristan Glatard

However, recent works have highlighted numerical stability challenges in DNNs, which also relates to their known sensitivity to noise injection.

Data Augmentation Through Monte Carlo Arithmetic Leads to More Generalizable Classification in Connectomics

1 code implementation • 20 Sep 2021 • Gregory Kiar, Yohan Chatelain, Ali Salari, Alan C. Evans, Tristan Glatard

The variability in the perturbed networks was captured in an augmented dataset, which was then used for an age classification task.

Age Classification Data Augmentation +1

Accurate simulation of operating system updates in neuroimaging using Monte-Carlo arithmetic

no code implementations • 6 Aug 2021 • Ali Salari, Yohan Chatelain, Gregory Kiar, Tristan Glatard

Overall, our results establish that FL accurately simulates results variability due to OS updates, and is a practical framework to quantify numerical uncertainty in neuroimaging.

Reducing numerical precision preserves classification accuracy in Mondrian Forests

1 code implementation • 28 Jun 2021 • Marc Vicuna, Martin Khannouz, Gregory Kiar, Yohan Chatelain, Tristan Glatard

Mondrian Forests are a powerful data stream classification method, but their large memory footprint makes them ill-suited for low-resource platforms such as connected objects.

Classification Human Activity Recognition