DOI: 10.22184/1993-7296.FRos.2026.20.5.372.387

Photonic neural networks are currently receiving increased attention due to the prospects of increasing computation speed and reducing energy consumption through the use of photons instead of electrons that in the future may solve the problem of slowing growth in the clock rate of digital processors (CPUs) and the challenge of reducing their power consumption. In turn, it will lead to the reduced data processing time in the fiber-­optic data network nodes and increased bandwidth of networks with the comprehensive structures. This article provides an overview of experimentally achieved specifications of nonlinear devices based on the common integrated photonics materials. The ultimately achievable characteristics of nonlinear photonic devices and the challenges facing their implementation are analyzed.
Analysis of the literary data and numerical estimates demonstrates that the devices based on semiconductor photonic crystal resonators have the greatest potential for nonlinear photonics.