Skip to content Skip to navigation

Synaptic rules and circuit architectures for learning from feedback

Stanford Neurosciences Institute, Jay Bhasin

In the language of engineering, the brain can be thought of as a feedback controller: given a goal, i.e., a desired state of the world, neural circuits in the brain generate motor output that alters the current state of the world, which is fed back into the system as sensory input, allowing the system to perform reactive, online correction during the time course of a behavior. However, many complex behaviors require rapid and precise timing, which is difficult to achieve by reactive control because of delays inherent in such a system. Instead, a predictive or feedforward controller is required, which uses the statistics of inputs and an internal model of the system being controlled to predict how a behavior will affect the state of the world, generating more precise, effective responses. Over the course of many responses, the brain can use feedback to learn; that is, to tune the internal model parameters themselves in an online manner through synaptic plasticity. In this project, I will use engineering principles from modern control theory, experimental neuroscience and computational neuroscience to significantly advance our understanding of how feedback driven plasticity in a tractable neural circuit–the circuit in the cerebellum and brainstem mediating learning in the vestibulo-ocular reflex–is orchestrated across multiple synaptic sites and over various timescales so that circuit dynamics are changed to improve performance.

Participants

Lead Researcher(s): 

Primary Advisor: Jennifer Raymond (Neurobiology)
Co-Advisor: Stephen Boyd (Electrical Engineering)

Funding Type: 
SIGF - Graduate Fellowship
Award Year: 
2018