Nina Miolane

This study uses persistent homology to analyze neural activity in brain organoids and finds structured topological features at the scale of typical recordings.

The paper introduces latent processing units (LPUs) to explain how neural networks perform reliable computations despite variability, using six theorems to unify geometric and dynamical perspectives.

The paper shows that low-dimensional neural coding suppresses noise and extends working memory duration in recurrent networks through theoretical analysis and experimental validation in mice.

The paper analyzes sex differences in brain networks across the lifespan using a new AI-based method, finding minimal differences in early childhood, pronounced in adulthood, and diverging in later life.

The paper introduces an unsupervised method to analyze structural variability in cryo-EM data by deforming atomic models using normal mode analysis directly against 2D data, improving accuracy and interpretability.

This paper frames the challenge of interpreting cryo-EM latent spaces as an Independent Component Analysis problem, aiming to develop identifiable models for conformational landscapes.

This study trains an RNN to path integrate trajectories of multiple agents, finding changes in neural representations compared to single-agent models.

This paper introduces DeCOr-MDS, a robust MDS method that detects and corrects orthogonal outliers to improve embedding quality in biological datasets.