St. Petersburg Researchers Develop Low-Cost Spin-Wave Neural Network Hardware
Researchers at St. Petersburg Electrotechnical University "LETI" have developed a magnonic system capable of performing neuromorphic computing using spin waves.
The device is built around a ceramic ferrite plate made of yttrium iron garnet, replacing the expensive single-crystal films used in conventional magnonic devices and sharply reducing production costs. Magnonics is considered a promising approach to hardware neural networks because the relatively low propagation speed of spin waves allows devices to be miniaturized.
In experiments, the system was tasked with recognizing digits from 0 to 9. At a noise level of 20 percent, accuracy reached 98 percent for large printed digits, exceeded 76 percent for smaller printed digits, and reached 71 percent for handwritten digits.
The research is being conducted under a government assignment from Russia's Ministry of Science and Higher Education with support from a mega-grant.
