On the effect of dimensionality reduction by Manifold Learning for Evolutionary Learning

Abstract

Manifold Learning has attracted much attention for this decade. One of the main features of Manifold Learning is that Manifold Learning tries to conserve local topologies in high-dimensional space. In this paper, we discuss the effect of the dimensionality reduction of input spaces of Evolutionary Learning. That is, we propose the use of Manifold Learning for Evolutionary Learning with redundant sensory inputs in order to avoid the difficulty of designing the allocation of sensors. The proposed method is composed of two stages: the first stage is to generate a mapping from higher dimensional sensory inputs to lower dimensional space, by using Manifold Learning. The second stage is using Evolutionary Learning to learn control scheme. The input data for Evolutionary Learning is generated by translating sensory inputs into lower dimensional data by using the mapping. We examine two Manifold Learning algorithms: Isomap and LLE. We adopt the Instance-Based Policy Optimization as an Evolutionary Learner. In addition, a metric of relative error of distances between original input space and reduced space is introduced. Experimental results of robot navigation problems show the effectiveness of the proposed method.