neuromorfe

Neuromorfe is a term used in neuroscience-inspired computing to describe neural units or networks that incorporate both the morphological structure of neurons and adaptive plasticity. In this usage, neuromorfe refers to architectures where the connectivity and sometimes the internal structure of the processing units can change over time in response to activity, akin to dendritic learning and synaptic remodeling. The concept sits at the intersection of neuromorphic engineering and computational neuroscience, and definitions vary across sources.

Unlike traditional fixed-topology neural networks, neuromorfe systems emphasize structural plasticity, multi-compartment models, and local learning rules that operate at the level of compartments or modules rather than single points. This enables context-sensitive integration of information, temporal pattern recognition, and energy-efficient computation.

Hardware and software realizations often employ multi-terminal synapses, memristive or phase-change devices, and configurable circuitry that can rewire or reweight connections, simulating morphological adaptation. In software, dynamic graphs and compartmental neuron models can capture dendritic-like computations and plasticity rules.

Applications include adaptive control, robotics, sensory processing, and research into brain-inspired learning. The field faces challenges such as hardware complexity, stability during structural changes, and the interpretability of learned morphologies. There is no single standard definition, and the term neuromorfe may be used differently depending on the author.

See also: neuromorphic engineering, morphological computation, plasticity, dendritic computation.