What if hardware itself could learn? This is the perspective explored in the scientific article “Self-organizing memristive networks as physical learning systems”, published in Nature Reviews Physics by an international research team including Dr. Gianluca Milano, Senior Researcher at INRiM and Associate Professor at Politecnico di Torino.
The study examines the potential of self-organizing memristive networks, systems composed of nanowires or nanoparticles connected through dynamic electrical junctions. When they receive a stimulus, these networks modify their conductive pathways and retain a memory of their previous activity.
This ability to change through experience makes them particularly promising for the development of new computing systems. Unlike conventional systems, in which artificial intelligence software is executed by a processor, in memristive networks the physical structure of the material itself directly contributes to information processing and learning.
One possible application is edge computing, in which data are processed directly where they are generated. In the future, sensors, robots, autonomous vehicles and satellites could analyse information locally, reducing energy consumption, response times and dependence on cloud computing.
The article is authored by Francesco Caravelli, Gianluca Milano, Adam Stieg, Carlo Ricciardi, Simon A. Brown and Zdenka Kuncic.
Gianluca Milano is also the principal investigator of the ERC Starting Grant MEMBRAIN, a project devoted to developing new computing architectures based on self-assembling materials. The project involves close collaboration between Politecnico di Torino and INRiM.