Abstract
The ability to identify characteristics of dynamical systems is important because it allows the understanding and modeling of complex physical phenomena that can be utilized in many applications, ranging from finance, environment, health, to industry. Here the concepts of convolution and pooling in deep learning are introduced to compute the eigenvalues of fuzzy recurrences obtained from the phase-space reconstructions of time series. Computer results show that the largest eigenvalues of convolved fuzzy recurrence plots can differentiate between cohorts of random and chaotic signals. Copyright (C) 2021 EPLA