We take a systems approach to studying the fundamental principles of computation in neural systems, combining data-mining methods with computational modelling.
To decode the information contained in experimentally-recorded neural activity, we are developing analysis methods that are able to take the recordings of simultaneous brain cell activity and automatically solve basic problems in understanding these recordings: finding when the cells are active together, which groups they belong to, and what form that co-ordinated activity takes.
To understand what these analyses imply about neural computation, we build models that explain the data, generate new hypotheses and make predictions. At an abstract level, we model how groups of neurons encode algorithms, and compute with them. At a biophysical level, we build detailed models of brain regions to study the dynamical repertoire of neural microcircuits, seeking clues for how different brain regions may be representing and computing information. We are applying these methods to problem areas in computation across neuroscience: the encoding of motor programmes, sensory coding, decision-making, and neural development.