Tracking Dynamical Transitions using Link Density of Recurrence Networks

Rinku Jacob,R. Misra,K P Harikrishnan,G Ambika

2024-05-24

Abstract:We present Link Density (LD) computed from the Recurrence Network (RN) of a time series data as an effective measure that can detect dynamical transitions in a system. We illustrate its use using time series from the standard Rossler system in the period doubling transitions and the transition to chaos. Moreover, we find that the standard deviation of LD can be more effective in highlighting the transition points. We also consider the variations in data when the parameter of the system is varying due to internal or intrinsic perturbations but at a time scale much slower than that of the dynamics. In this case also, the measure LD and its standard deviation correctly detect transition points in the underlying dynamics of the system. The computation of LD requires minimal computing resources and time, and works well with short data sets. Hence, we propose this measure as a tool to track transitions in dynamics from data, facilitating quicker and more effective analysis of large number of data sets.

Data Analysis, Statistics and Probability

What problem does this paper attempt to address?

This paper proposes a new method for tracking transitions in dynamic systems using Link Density (LD), especially for complex systems based on time series data. The author demonstrates LD as an effective indicator for detecting dynamic transitions using the time series data of the Rössler system, which can effectively identify bifurcation transitions and dynamic change points during slow parameter changes.

In bifurcation transitions, the paper indicates that LD sharply decreases during the initial period doubling bifurcation and continues to decrease with subsequent bifurcations until the system enters a chaotic state. In each periodic window, LD rises again when transitioning from a chaotic state back to periodicity. Additionally, the standard deviation (STD) of LD increases in the chaotic region, indicating a more pronounced change in the network structure, which helps identify transition points between periodic and chaotic states.

In the case of slow parameter changes, the paper applies sliding window analysis to track the dynamic changes of the Rössler system. Even when the parameter change rate is much slower than the system's dynamical time scale, LD and its STD effectively detect transition points in the system.

In conclusion, this study emphasizes LD as a simple and effective method for identifying dynamic changes in complex systems, particularly suitable for handling limited datasets and real-time applications. The paper suggests using the STD of LD as an early warning signal to rapidly and efficiently analyze dynamic transitions in large datasets.