CNN을 활용한 카오스 신호 분류 검증

초록

The aim of the study was to classify the chaotic time-series data with the nonlinear problem using the convolutional neural network (CNN), and to determine and verify the chaotic characteristics from a deterministic system. The classical nonlinear differential equation established by the Rossler model was used, and the chaotic characteristics were determined by the Lyapunov exponent. The chaotic properties was visualized using an unthresholded recurrence plot through the proposed procedure. A simple CNN model was developed to learn the extracted image using the proposed feature-visualization technique. As a result, the chaotic characteristics were classified with an accuracy of 99% or more.

키워드