The last two decades have seen an explosion in our ability to collect and analyze complicated, high-dimensional neural datasets. For instance, using advance imaging or high-density multi-channel electrode arrays, we can now examine activity from thousands to hundreds of thousands of individual neurons, simultaneously. What we have learned from these recordings is that the brain is not static and it is not random: neurons change patterns of activity dynamically and exhibit intricate patterns of correlation. We also know far more about the components of neural circuits than ever before. Perhaps one of the best exemplars of this rich structure are the constellations of cell types inferred from patterns of single-cell RNAseq expression. Where once a modeler would have constructed a network of excitatory and inhibitory cells, we now know fine details of microcircuity composition and connectivity. Despite the wealth of detail, there is nowhere near enough data to exactly circuits in the level of detail necessary if one even wanted an exact model of the brain. Moreover, we know, from fundamental work in neuromodulation, that it is not good enough to know network parameters on average or one at a time. Depending on relationships between parameters, networks with the same average parameters can show fundamentally different behavior.
What biological details matter? What aspects of neural activity patterns are surprising? How do we know what is surprising? The Sederberg group uses computational and theoretical methods to examine these questions across many systems — different brain areas, different individuals, and different species. We are a theory-driven lab, but we engage deeply with data and experiment through extensive collaboration.