A clear and important signature of neural response is the large degree of variability from trial-to-trial in sensory and motor tasks. Neural variability is malleable depending upon task specifics, cognitive state, and sensory input; however little is known about the mechanisms that mediate this variability. Using techniques from non-equilibrium statistical mechanics and nonlinear systems theory this project will build a coherent theory of the mechanics of neural variability. Recent experiments across cortex show that the stochastic dynamics of spontaneous neural activity is very rich, beyond that observed during evoked or task driven responses. This research will show how clustered neural architectures can replicate this finding. This research will develop key insights from simplified Markov chain models of cortical assembly dynamics, which will allow a more formal approach to our previous simulation based studies. Uncovering the core mechanics of neural variability is a critical step in giving a foundation for a theory of neural computation.
Modern computers minimize noisy fluctuations in transistors and semiconductors in order to improve performance reliability. In stark contrast, brain dynamics show a sizable trial-to-trial variability of neural responses, making it clear that the nervous system works under different principles than silicon machines. This research will use contemporary techniques from statistical mechanics and nonlinear system theory to give a theoretical foundation to the mechanics of neural variability. Specifically, our theory will expose how the underlying circuitry of the brain is involved in producing and controlling neural variability. This research will guide future experiments that aim to better characterize neural circuits, placing any data in the context of a core functional feature of the nervous system. Furthermore, our research will give critical insights concerning many neurodegenerative diseases where a common neural signature is an excess of variability.
