Uncovering neural algorithms for learning

Performing an ace tennis serve or checkmating an opponent in chess can seem like daunting tasks for a novice. Yet we take it for granted that, given enough practice and time, we can all become expert practitioners of such skills. The brain’s ability to solve complex learning challenges is an incredible feat whose speed and efficiency is unmatched by machine intelligence. However, little is known about this ability and the neural circuits that underlie it.

We employ an interdisciplinary strategy to investigate the neural basis of skill learning in rodents. Using a fully-automated behavioural training system we acquire large datasets as animals learn to solve complex motor and foraging tasks. We infer trial-by-trial learning strategies by performing detailed analysis of massive behavioural datasets in concert with computational modelling. Using targeted perturbations of neural circuits and an experimental platform to automatically record the spiking activity of large ensembles of neurons continuously (24/7) over weeks and months in behaving animals, we investigate how learning strategies are implemented by brain circuits.