Neuroncentric

Neuron-centric is a term used in neuroscience and artificial intelligence to describe approaches that treat neurons as the primary units of information processing. It contrasts with views that prioritize networks, architectures, or algorithms, and emphasizes biological plausibility and local computation within individual neurons and their connections.

In neuroscience, neuron-centric models view neurons as nonlinear integrators of synaptic inputs that produce spikes when a threshold is exceeded. Classic models include integrate-and-fire and Hodgkin-Huxley neurons. Learning is often conceived as local synaptic plasticity, such as Hebbian rules or spike-timing-dependent plasticity (STDP), shaping connectivity based on activity.

In AI and neuromorphic engineering, neuron-centric design motivates spiking neural networks and hardware that emulate neural elements. Such systems aim for energy-efficient, event-driven computation and temporal processing by implementing neurons and synapses in silicon. Training can combine local learning rules with global signals or use surrogate gradients to tackle optimization.

Critics contend that focusing solely on neurons can miss emergent properties that arise from large-scale networks and multi-scale interactions. Proponents argue that neuron-centric approaches provide a tractable, biologically grounded framework for understanding brain function and building efficient neuromorphic systems.

Related topics include neuron, spiking neural network, neuromorphic engineering, Hebbian learning, and spike-timing-dependent plasticity.