Constraint Satisfaction Problems solution through Spiking Neural Networks with improved reliability: the case of Sudoku puzzles

Solutions to constraint satisfaction problems can be found by means of spiking neural networks through a stochastic evolution following attractor dynamics. An example of CSP is represented by the latin square problem, whose subclass of Sudoku puzzles is accounted for in this work. Starting from a previous implementation available in the Literature, we propose a modified SNN-based solver which offers a fully spiking approach and improved reliability with respect to the problem definition.