Cognitive Computational Neuroscience Group

The group asks which features of biological organization matter for function, meaning what the brain can do and how it does it. We approach this question from two directions.

From the side of biology, we build biophysics-informed deep learning models and functional whole-brain models. In these models, features of biological structure can be imposed, varied or removed. The features range from retinal sampling and cortical layers and columns to large-scale connectivity. Each manipulation shows what a feature contributes to what the model can do.

From the side of function, we train models to perform tasks and then ask how much they come to resemble the brain. AngoraPy, our toolbox for goal-driven training of brain-inspired models, and our neural network model of dexterous hand control take this route.

We compare models with human behaviour and neural dynamics. Where their predictions call for it, we run psychophysical experiments. Our work spans visual content that is presented externally and content that is generated internally. Applications include how the visual system adapts to and recovers from disorders such as glaucoma, and face perception in forensic settings such as identifying AI-enhanced imagery.