Abstract
The purpose of this study is to explore advanced methods of complex system dynamics to discover latent patterns from nonlinear time series of personalized major depression. The study was performed with methods for analysis of complex system dynamics, including fuzzy recurrence plots, fuzzy joint recurrence plots, fuzzy weighted recurrence networks, and tensor decomposition of the recurrence dynamics. Both the use of two complex network properties known as the average clustering coefficient and characteristic path length and the tensor decomposition of the fuzzy weighted recurrence plots of the depression time series suggest a critical transition as an early warning signal in the reduction of anti-depressant medication applied to a single participant. Fuzzy recurrence plots, fuzzy recurrence networks, and tensor decomposition of mental-state dynamics are useful mathematical tools for constructing patient-specific models of the dynamics of depression and detecting development of new depressive episodes over the effect of drug-dosage alteration.