Laser Network as an Optimization Machine


  Sagie Gadasi  ,  Igor Gershenzon  ,  Geva Arwas  ,  Chene Tradonsky  ,  Asher. A. Friesem  ,  Oren Raz  ,  Nir Davidson  
Weizmann Institute of Science

Recently, there has been growing interest in the utilization of physical systems as heuristic optimizers for classical spin Hamiltonians. Gain-dissipative optical oscillator networks have been suggested and explored for this purpose. Such networks inherently suffer from the intensity heterogeneity effect which results in an inexact mapping between the oscillator network loss rate and the spin Hamiltonian. We theoretically analyze and experimentally demonstrate a scheme for the solution to the intensity heterogeneity problem in a laser network classical XY model simulator based on a degenerate cavity laser. We find that for each XY model energy minimum there corresponds a unique set of laser loss values that leads to a laser network fixed point with uniform intensity and to phase values that coincide with the XY model minimum. We find that without proper intensity equalization, the laser network phase deviates from the XY model minimum state phase values. Moreover, an inherent contradiction exists between finding the ground state close to the oscillation threshold and accurately mapping the network state to an XY state due to the significant intensity non-uniformity close to the network oscillation threshold.In the example laser network studied in this work, the deviation from the minimal XY energy was improved by a factor of 8 by employing the intensity equalization scheme.