The circuitry of the human brain is stunningly complex, with billions of neurons linked by more connections (synapses) than there are stars in the Milky Way. Neurons communicate via tiny electrical impulses called spikes, exchanging over 100 billion spikes per second. It is often thought that, to ensure robustness against noise, a single spike must play only a negligible role, and information must be encoded in more complex sequences. Yet our brain is capable of processing very weak signals, such as the detection of few photons of light. How is such high sensitivity compatible with noise rejection?
Writing in Physical Review X, Michael Monteforte and Fred Wolf at the Max Planck Institute for Dynamics and Self-Organization in Germany study a simplified model of cerebral circuits, based on an ensemble of coupled oscillators. They analyze how the network responds to single spikes as well as to fluctuations of other variables, such as membrane potentials (electrical potentials across the membrane of a neuron cell).
The authors’ simulations show that the network behaves as if its parameters are moving in a space made of many stability domains, termed “dynamic flux tubes.” In each domain, the network evolves along a stable and predictable trajectory. While small fluctuations and noise cannot drive it out of a stable domain, single spikes can switch the system from one domain to another and set the entire neural network on a completely different path. In other words, one single neuronal spike in a sequence of billions can affect the response of the entire network.
These results may help explain how our brain separates noise from true signals and find application in the design of artificial neural networks with tailored response to different types of stimuli. – Matteo Rini