Main Menu (Mobile)- Block
Main Menu - Block
Why is the brain organized as it is? What are the general principles that govern neural computation? We study the strategies by which information is processed in neural populations, with the goal of connecting between the sensory environment and behavior.
How does collective functionality emerge from individual system components?
This fundamental question manifests itself in a variety of contexts familiar to physicists, engineers, and neuroscientists alike, where it takes forms as diverse as how properties of a material arise from individual particles, how different design architectures shape information flow in engineered circuits, or how cognitive function emerges from neuronal networks.
Evolution has been seeking answers to these same questions for the past several billion years, and we are witnesses to its variety of intermediate successes. As the products of such successes, biological systems have realized a host of creative solutions that reflect both physical and metabolic limitations of biological building materials, as well as external limitations imposed by the statistical regularities of the natural world. Yet the very aspects of biological systems that enable them to realize such diverse organizations also make them inherently difficult to study: they are built from a collection of both optimized and outdated components, these components adapt and degrade over time, and the environment to which these components adapt is itself changing over time.
What are the solutions that the brain has realized for performing neural computations in the midst of such variability? And why are these solutions unlike those we might use to design a computer? Our work explores the nature and functional advantages of neural design solutions, with the goal of uncovering general principles that govern information processing in the brain. To this end, we develop theoretical, computational, and data-driven tools to probe how different cellular architectures support and constrain collective function in neural populations. And because the functionality of a system is intimately tied to the structure of its environment, we try to understand neural design principles in the context of natural signals and efficient behavioral strategies.
"If you just have a single problem to solve, then fine, go ahead and use a neural network. But if you want to do science and understand how to choose architectures, or how to go to a new problem, you have to understand what different architectures can and cannot do."