Learning to code in a noisy cortex
Abstract
A critical function of the brain is to learn and store the important associations that make up our world. However, a fundamental reality of the brain is that neurons and synapses operate in the face of a tremendous amount of noise, both external and internal to networks. On the surface, reliable storage and neuronal variability would seem to be in opposition of one another. In this proposal we outline a combined theoretical and experimental strategy to develop and test a principled theory for how biological plausible activity dependent synaptic plasticity gives rise to strongly coupled assemblies of neurons in cortical networks. The key advance is that our theory will not only capture the deterministic aspects of learning, but also a learning dependent population-wide variability. With an account of neuronal variability in learning we will explore how neuronal plasticity shapes neural coding, where the response ‘noise’ is just as prominent as the response signal. We will show how learning-based population variability is a reflection of a mechanism that stabilizes, as opposed to dissolves, assembly structure. We will test our theory with novel experiments in rodent orbitofrontal cortex that both measure the plasticity mechanisms that support assemblies as well as targeted manipulations of assembly formation itself. Past theories of neuronal learning have suffered from a lack of biophysical realism, obscuring how results inform how actual synaptic plasticity shape neural computation. Our proposal will bridge theories of synaptic plasticity with theories of neural variability so as to build a more complete picture of cortical processing.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Nov 06, 2017
- Source ID
- N000141812002
Entities
People
- Brent Doiron
Organizations
- Office of Naval Research
- United States Navy
- University of Pittsburgh