The DePasquale lab develops mathematical models to understand how populations of neurons perform computations to produce behavior. Broadly, we take two approaches. One is data-driven: we collaborate with experimental neuroscientists to develop tailored machine learning models of neural activity to identify the algorithms that drive behaviors such as decision-making or movement. Our second approach is theoretical: we construct and analyze artificial neural network models to understand how their structure gives rise to analogous computations and other functional features observed in biological neural circuits. A key area of interest has been understanding how structured connections between neurons produce network activity that is both coherent and irregular, a common yet paradoxical feature of neural responses.<\/p>\n