Natural computing in the brain: geometry, criticality, and fusion.

点击翻译按钮获取中文摘要

Nature computes, and nowhere is this more evident than in the primate brain. Understanding the brain holds the key to understanding human behaviour, cognition, and equally significantly, to better understanding the nature of computation itself. Indeed, the current neural network AI renaissance is powered by models inspired by computation in the brain. But the brain is one of the most complex systems ever studied by science. My research will attack this complexity through the formalized framework of Natural Computing and adoption of new mathematical tools. Contemporary neuroscience studies the brain as a network; this is an apt metaphor for a complex computing machine consisting of many distinct, communicating, subunits. My team will build on the success of the network metaphor by extending models of the brain with approaches based on self-organized criticality and data fusion. Dynamical systems in a critical state display immense computational complexity. Critical states are the knife edge between boring, homogenous, stasis, and unmanageable chaos. Neuroscientists have been studying the "critical brain hypothesis" for decades and there now exists compelling supporting experimental evidence. Starting with the brain networks already richly described in the literature, my team will study computational dynamics on these networks using the framework of self-organized criticality and new tools from tropical geometry. These models will offer new insights into the nature of neural computation, permit us to answer questions about the optimality of observed brain networks, and also drive forward the basic mathematical study of self-organized criticality on complex networks extracted from nature. But are all nodes in brain networks equal? Evolution has created brains composed of distinct subunits; the structure, function, and connectivity of, e.g., the brainstem, is profoundly unlike that of the prefrontal cortex. Nevertheless, the brain appears to compute as a single, integrated, unit. In a world of unbounded sensor growth, and multiscale computational capacity, computer scientists and engineers have developed the tools of data fusion to deal with exactly this challenge of integration. My team will also take a data fusion approach to modelling communication in brain networks inferred from high-resolution neuroimaging. This will provide a framework for understanding the complex co-ordination, and communication, that takes place during neural computation. I have two decades of experience in the field of natural computing, and nearly a decade of experience in neuroscience, having used a sabbatical leave in 2011 to earn an MSc in neuroscience. I am a core member of Western's Brain and Mind Institute, a faculty affiliate of the Vector Institute for Artificial Intelligence and an associate scientist at the Lawson Health Research Institute. My team has the experience, expertise, environment, and established collaborations to pursue this work.