Locally Linear Embedding Based on Soccer Match Algorithm in Its Optimization Process

Manifold learning is an important scientific research branch of artificial intelligence (AI) related algorithm. Its main purpose is to realize nonlinear data dimensionality reduction whilst keeping the characterization between original data points. One of the most representative manifold learning algorithm is locally linear embedding (LLE), its main idea is to realize algebraic isomorphism of data from high-dimensional space to lowdimensional space, and it is essentially a 2-stage optimization problem, moreover, under the premise of using L1 norm in its objective function without amplifying or reducing errors, it is worth to explore the effective optimization algorithm for it. Considering that a specific data point is represented by a finite number of surrounding data points in space, and inspired by the swarm intelligence algorithm called soccer match algorithm (SMA), in this paper, LLE based on SMA in its optimization process has been proposed, in which the optimization of decision variables for a finite number of soccer players is correspond to the optimization of parameters for a finite number of data points. Via numerical experiment, better optimization applicability of the proposed algorithm has been demonstrated.