Comparing Particle Swarm Optimisation and Genetic Algorithms for Nonlinear Mapping

Reducing the dimensionality of high-dimensional data simplifies how data is presented, allowing easier visualisation of high-dimensional data and facilitating more efficient extraction of knowledge. Nonlinear mapping methods transform data existing in high-dimensional space into a lower-dimensional space such that the topological characteristics of the high-dimensional data are preserved. Recent work proposed a particle swarm optimisation algorithm to perform nonlinear mapping. This paper compares a number of optimisation algorithms in performing nonlinear mapping. Experimental results distinguish between each of the optimisation algorithms. Nonlinear mapping methods were designed to map small datasets and are unable to project new data points. A proposed method to perform nonlinear mapping on large datasets is discussed.