Metasurface network learns without a digital model
A Nature Communications paper published on June 24, 2026 describes a physical neural network with four programmable metasurfaces, trained through electromagnetic interactions to focus waves, recognize objects, and perform localization without digital models.
A Nature Communications paper published on June 24, 2026 presents a physical neural network made of four programmable metasurfaces. It describes a training method that dispenses with digital models: the network learns through electromagnetic interactions to focus waves, recognize objects, and perform localization.
The finding concerns physical computing, not a Rota Nacional feature. To check the methods and limits of the claims, consult the original paper in Nature Communications and compare this account with its full text and data. If you use AI to study or apply the work, avoid sending personal data or internal information unless necessary, and follow your organization's privacy policies.