Discovering how brain structure shapes function
Overview
Brains are composed of hundreds to billions of neurons, and complex structures at multiple scales. From the intricate shapes of single neurons to the pathways which define whole-brain connectivity.
Our research combines biological data analysis with artificial neural network models to discover how these structures sculpt our perceptions, thoughts and actions (brain function).
Understanding this is important as it would allow us to better understand disease processes - in which structural changes lead to dysfunction, and to build better machine learning models - by taking inspiration from the brain.
More technically
We study how the structure of a neural network shapes its function. With a focus on multisensory integration.
To do so, we develop computational models with biologically-inspired structural features. For example, sparse bidirectional connectivity, or modularity. Then we:
- Compare their function by modelling them as artificial neural networks and training them to perform psychophysical or naturalistic tasks. To do so, we use machine learning algorithms (including supervised and reinforcement learning) and design new tasks and simulated environments.
- Aim to discover why different structures function differently. In this direction, we use methods from neuroscience, dynamical systems and machine learning interpretability, and develop new approaches.
Finally, to test our structure-function hypotheses we collaborate with experimental research groups and make use of open-source datasets.