We understand the world in a strikingly flexible manner. We can perceive objects in a way that is highly specific, or general and abstract, we can be sensitive to specific features, or completely disregard them. Importantly, we can switch between these modes in an eye blink. How does the brain achieve this?
We are looking to answer this question by studying the patterns of neural activity in the neocortex during visual perception. These activity patterns not only consist of a large number of neurons (are high-dimensional), but are also constantly changing due to recurrent network interactions, giving rise to intricate dynamics. Making sense of such intertwined high-dimensional networks is notoriously difficult. Despite its complexity, the neocortex is highly structured, organized into anatomically defined areas, layers, and cell types. This organization is a helpful guide in uncovering its computational principles. For example, we can study how long-range cortico-cortical projections or local excitatory-inhibitory balance coordinate specific dynamical modes.
To this end, we employ a combined experimental and computational approach. We use distributed, multi-region high-density electrophysiological recordings of neural population activity in the visual cortex of mice, in conjunction with transient optogenetic perturbations, to dissect the functions of various features of network connectivity. In parallel, we use analytical tools from dynamical systems analysis to distill our observations into coherent models. As such, we iterate between developing hypotheses for neural computation and experimentally validating them.