We analyze and model the joint activity patterns of large populations of neurons, asking how they encode information, how we may read it, and how the brain decodes it.
Using different families of statistical models that were highly accurate in describing the codebook of neural systems in different organisms, we idenditied simplifying principles that govern the nature of the code, how to learn neural dictionaries and thesaurus that allow us (and would allow the brain) to decode novel patterns. These models suggest a normative theory of the computations that the underlying circuits perform.
Learning such accurate models would allows us to uncover the computations that neural circuits perform, to build better and more efficient brain machine interfaces, and ultimalty not only read information from the brain, but also write it in.