I am a Lecturer in Data Science for Life Sciences at the University of Edinburgh.

Modelling neural activity with machine learning and statistics

In our group, we focus on developing flexible probabilistic and machine learning methods for modelling and analysing neural activity.

Deep learning models for predicting brain activity

We developed a Vision Transformer model to predict primary visual cortex activity based on visual stimuli and behaviour.

Model block diagram and synthetic receptive fields.

• B. Li, I. M. Cornacchia, N. L. Rochefort, A. Onken (2023). V1T: large-scale mouse V1 response prediction using a Vision Transformer. Transactions on Machine Learning Research 2835-8856.

Probabilistic models of neural relationships

We use techniques like copulas, Gaussian processes and normalizing flows to model varying interactions within neural activity and their correlation with external variables.

Copula dependencies between two example neurons (top) and a neuron and the lick variable (bottom) changing over time (left to right).

• N. Kudryashova, T. Amvrosiadis, N. Dupuy, N. Rochefort, A. Onken (2022). Parametric Copula-GP model for analyzing multidimensional neuronal and behavioral relationships. PLoS Computational Biology 18(1):e1009799.
• L. Mitskopoulos, T. Amvrosiadis, A. Onken (2022). Mixed vine copula flows for flexible modeling of neural dependencies. Frontiers in Neuroscience 16:910122.

Matrix and tensor factorizations for neural dimensionality reduction

We decompose neural matrix and tensor representations to extract interpretable structure from large datasets.

Decomposition of a tensor representing neural population activity into a small core tensor and several matrices.

• L. Mitskopoulos, A. Onken (2023). Discovering low-dimensional descriptions of multineuronal dependencies. Entropy 25(7):e25071026.
• T. Tsunematsu, A. A. Patel, A. Onken, and S. Sakata (2020). State-dependent brainstem ensemble dynamics and their interactions with hippocampus across sleep states. eLife 9:e52244.

Opportunity to do a PhD

If you would like to join my lab as a PhD student, please contact me and tell me something about your background and research interests. Funding opportunities available!