Control of nonstationary multimode lasing by neural networks with radius basis functions

A neural network with radius basis functions is trained by using the evolution approach to recognise the mode characteristics of a Fabry–Perot resonator. It is shown that the network is capable of determining simultaneously the qualitative and quantitative characteristics of nonstationary lasing. The efficiency of the approach for determining the amplitudes of the fundamental and first modes on passing from single-mode to two-mode lasing is demonstrated by the example of a network with the response time shorter than the resonator round-trip transit time. The efficiency of a new scheme, in which field-amplitude detectors are located inside the resonator outside of the output aperture, is also demonstrated.