All-optical Sudoku solver with photonic spiking neural network

Abstract We proposed a Sudoku solver based on photonic spiking neural network (SNN) consisting of vertical-cavity surface-emitting lasers with an embedded saturable absorber (VCSELs-SA). Here, N 2 neurons are adopted to solve the N × N Sudoku puzzle with a unique solution. Values in Sudoku are temporally spike-encoded in different spiking times. The result is also achieved according to the spike timing information. In a network, for these VCSELs-SA belong to the same row or column or quadrant, they are inhibitory connected., i.e., from X polarization (XP) mode to Y polarization (YP) mode. Otherwise, they are excitatory connected, i.e., from XP mode to XP mode. Polarization mode competition (PMC) mechanism of VCSEL-SA is utilized to ensure the priority of inhibition behavior. With numerical simulations, 3 × 3 Sudoku puzzle is successfully solved with the help of inhibitory dynamics of VCSEL-SA. The results are represented by the output spike timing of XP mode of each VCSEL-SA. Furthermore, the photonic SNN is also extended to solve 4 × 4 Sudoku puzzle. This work may be interesting for the brain-inspired photonic neuromorphic information processing.