A meta-cognitive interval type-2 fuzzy inference system and its projection based learning algorithm

• DOI: 10.1007/s12530-013-9102-9
• Publication Date: 2013-12-30T15:07:42Z
• Authors: Kartick Subramanian; Ankit Kumar Das; Suresh Sundaram; Savitha Ramasamy
• Keywords: Component (thermodynamics)

A meta-cognitive interval type-2 neuro-fuzzy inference system (McIT2FIS) based classifier and its projection based learning algorithm is presented in this paper. McIT2FIS consists of two components, namely, a cognitive component and a meta-cognitive component. The cognitive component is an interval type-2 neuro-fuzzy inference system (IT2FIS) represented as a six layered adaptive network realizing Takagi-Sugeno-Kang type inference mechanism. A self-regulatory learning mechanism forms the meta-cognitive component. IT2FIS begins with zero rules, and rules are added and updated depending on the prediction error and relative knowledge contained the current sample. As each sample is presented to the network, the meta-cognitive component monitors the hinge-loss error and class-specific spherical potential of the current sample to decide what-to-learn, when-to-learn and how-to-learn them, efficiently. When a new rule is added or when an existing rule is updated, a projection based learning algorithm computes the optimal output weights with least computational effort by finding analytical minima of the nonlinear energy function. It uses class specific criterion and sample overlap criterion to estimate the network parameters corresponding to the minimum energy point of the error function. Moreover, consistently under - performing rules are pruned from the network leading to a compact network. The performance of McIT2FIS is first evaluated on a set of benchmark classification problems from UCI machine learning repository. A tenfold cross validation based performance comparison with other state-of-the-art approaches indicates its improved performance. Next, its performance is evaluated on detection of attention deficiency hyperactivity disorder (ADHD) in children. The aim of this study is to classify a child as having typically developing controls or as an ADHD patient. Voxel based features extracted from amygdala region of the brain is employed in this study. The network is trained and tested on samples obtained from ADHD-200 consortium dataset consisting of 941 subjects. The performance comparison with standard support vector machine shows that McIT2FIS has superior classification ability than SVM in diagnosing ADHD.