Dr Timothy O’Leary, Dept Engineering, University of Cambridge
Abstract
Neurons and neural circuits have lots of parameters that are subject to biological feedback control. Often, multiple components affect a relatively simple, feedback variable that is then used to control all components collectively, a scenario we call degenerate feedback. In single neurons, receptors and ion channels shape average electrical activity, and this activity itself regulates expression of receptors and channels over slow timescales. In neuronal networks, synaptic connections are shaped by reward signals during learning, while the connectivity of the network determines performance, and therefore reward. In both cases there are nonunique configurations that solve a given task or objective, which makes it difficult to relate neural function to behaviour in general. I will describe recent work that explores the consequences of degenerate feedback control in the context of learning and representing the external environment. We find that noisy feedback on task performance can enable a network to learn, and redundancy in the network connectivity can make a fixed task easier to learn in the presence of an imperfect learning rule.