A neuroevolutionary method for knowledge space construction

Abstract

In this paper we propose a novel method for the construction of knowledge spaces based on neuroevolution. The main advantage of the proposed approach is that it is more suitable for constructing large knowledge spaces than other traditional data-driven methods. The core idea of the method is that if knowledge states are considered as neurons in a neural network, the optimal topology of such a neural network is also the optimal knowledge space. To apply the neuroevolutionary method, a set of analogies between knowledge spaces and neural networks was established and described in this paper. This approach is evaluated in comparison with the minimized and corrected inductive item tree analysis, de facto standard algorithm for the data-driven knowledge space construction, and the comparison confirms the assumptions.