Use of neural-network systems to control transient multimode lasing

A three-level neural network is considered which contains twenty five inputs, two hidden elements, and four outputs and is trained to recognise four situations at the input: the appearance of the fundamental mode of a Fabry–Perot resonator, the superposition of the fundamental and the first mode of the resonator with zero phases, the superposition of the fundamental and first mode of the resonator with phases 0 and π, and the appearance of the second mode and the superposition of the fundamental and second modes. It is shown that the network can recognise correctly variations in the mode composition of the Fabry–Perot resonator over the modulus of the transverse distribution of the field during the development of two-mode lasing. The operation of the network is studied for the two response times: the time shorter than the resonator time and the time determined by the response time of a pyroelectric detector of laser radiation.