Using Neural Networks in the Search of Low Auto-correlation Binary Sequences

Binary sequences with low auto-correlation are used in various applications, ranging from telecommunications and radar technology to mathematics and physics. Finding such binary sequences is a difficult optimization problem, because of its huge search space and many global and local optima. To mitigate this, researchers often use a skew-symmetry approach that significantly reduces the search space size. In this paper, we present a novel approach that utilizes a neural network. The neural network decreases the search space size to the same extent as the skew-symmetric approach, effectively guides the search algorithm, and consequently improves the solver. This approach is tested on binary sequences of lengths 31 and 41, respectively. We evaluate it with two different neural networks and compare them to the approach that only uses the skew-symmetry. Experimental results show that our approach requires fewer steps to find a solution of equal quality in comparison with the skew-symmetric approach.