A neural network that has been trained to detect temporal correlation and distinguish chaotic from stochastic signals.
The '/auto_ANN_Omega/' directory depicts the fully automatic code with the necessary libraries.
The main file 'chaos_detection_ANN.py' contains all the information.
The files:
W1.dat', 'W2.dat', 'B1.dat', 'B2.dat' are the weights of the ANN.
'colorednoise.py' is the library to generate the flicker noise (colored noise).
Instructions for running the code:
python chaos_detection_ANN.py serie.dat
'serie.dat' is the time series to be analyzed.
The code compares the time-series with 1 flicker-noise time-series with the same correlation coefficient (predicted by the ANN)
and the same length.
For small time-series length<1000 points we suggest the command:
python chaos_detection_ANN.py serie.dat 10
In this case, the code compares the time-series with 10 flicker-noise time-series.
The 'tests' directory presents an autorun of the Figure 3 for a practical use, with fewer points (101 initial conditions insted of 1000) and less precision (length 2^16 instead of 2^20).
Instructions for running the code:
python autorun3.py
After a few minutes the figure 'test_fig3.png' is generated.