On the Accuracy-Convergence Tradeoff in Sigmoid Fuzzy Cognitive Maps

Abstract

Recently, a learning procedure to improve the convergence of sigmoid Fuzzy Cognitive Maps was proposed. This algorithm estimates the slope of each sigmoid neuron while preserving the causal weights. This paper proposes a more realistic error function for this algorithm, which is based on i) the dissimilarity between two consecutive responses, and ii) the dissimilarity between the current output and the expected one. As a second contribution, we introduce four sufficient conditions to arrive at stability features. These conditions allow assessing the accuracy-convergence tradeoff attached to the proposed learning procedure.