A comparative evaluation of the Generative Topographic Mapping and the Elastic Net for the formation of Ocular Dominance stripes

In this paper we compare the self organising capabilities of the Generative Topographic Map (GTM) [1] and Elastic Net (EN) [2]. We analytically compare the two algorithms and examine the different ways in which they preserve topography by considering their respective ‘state space trajectories’. We present simulations that demonstrate the differences between the two algorithms. We conclude by using the GTM to simulate the formation of Ocular Dominance (OD) stripes and compare against earlier simulations using the EN. Our findings indicate that the GTM produces patterns with some of the required characteristics and match results obtained with the EN to a degree.