Neural Heterogeneity Controls Computations in Spiking Neural Networks

What is it about?

Neurons are the basic information-encoding units in the brain. In contrast to information-encoding units in a computer, neurons are heterogeneous, i.e. they differ substantially in their electrophysiological properties. How does neural heterogeneity affect the function of neural circuits? We derive and analyze a mathematical model of networks of heterogeneous spiking neurons, and show that neural heterogeneity affects excitatory and inhibitory neurons differently. In inhibitory neurons, heterogeneity controls whether the inhibitory population preserves or overwrites the dynamic regimes of the excitatory population it interacts with. In excitatory neurons, heterogeneity controls important computational properties such as memory capacity and function generation capacity. Our results suggest that homogeneous and heterogeneous networks can both be optimally tuned for specific and distinct functions in the brain.

Featured Image

Photo by Daniel Öberg on Unsplash

Photo by Daniel Öberg on Unsplash

Why is it important?

To understand how functions such as the precise control of body movements or the ability to understand and produce meaningful speech signals arise from networks of interacting neurons in the brain, mathematical models of these networks are essential. However, despite neurons in the brain being inherently diverse, most mathematical models of neural networks have neglected this diversity and instead focused on networks of identical neurons. In this work, we offer a mathematical model that elegantly incorporates neural diversity (or heterogeneity, as we call it) and we provide detailed insight into how this diversity affects the behavior of the model. We expect this to be an important contribution to developing neural network models of brain function that acknowledge one of the most pronounced features of neurons - their diversity.