Cognitive Assessment of Scientific Creative Skill by Brain-Connectivity Analysis Using Graph Convolutional Interval Type-2 Fuzzy Network

Abstract

Scientific creativity refers to natural/automated genesis of innovations in science, propelling scientific, technological, industrial, and/or societal progress. Mental paper folding (MPF) requires spatial reasoning, which is an important attribute to determine creative potential of people. The article proposes a novel approach to determine creative potential of people from their brain-connectivity network (BCN) during their participation in MPF tasks using functional near-infrared spectroscopy (fNIRS). The work involves three phases. The first phase includes construction of BCN using Pearson's correlation method. The centrality features of the nodes in the network are assessed in the second phase and transferred to a proposed graph convolutional-interval type-2 fuzzy network (GC-IT2FN) in the third phase to classify the creative potential of individuals in four grades. The novelty of the work includes: 1) a novel self-attention mechanism in the network to guide graph convolution layers to focus on the most relevant nodes; 2) selection of a new activation function, Logish, after graph convolution to enhance classifier accuracy; and 3) utilizing the promising region in the footprint of uncertainty (FOU) of the used fuzzy sets of IT2FN-based classifier to reduce the effect of uncertainty in brain data on classifier performance. Experiments conducted demonstrate the efficacy of the proposed framework in contrast to traditional approaches.