Implement Autoreservoir computing for multistep ahead prediction based on the spatiotemporal information transformation with Python. The code refers to the matlab program of the original author's repo.

If you find it useful, you are welcome to click Star in the upper right corner. Thank you very much.

First, install the necessary dependent packages:

pip install -r requirements.txt

For Lorenz model simulation, there are the following three cases:

• noise-free & time-invariant case: use mylorenz.py to generate high-dimensional data, set noisestrength = 0 in main.py;

• noisy & time-invariant case: use mylorenz.py to generate high-dimensional data, set noisestrength to be 0.1-1.0 in main.py, respectively;

• time-varying case: use mylorenz_dynamic.m to generate high-dimensional data, set noisestrength = 0 in main.py.

The code LongerPredictionSamples_ARNN.py in repository can generates the results in Figure 2d,2e,2f of the main text.

Expected running time for this demo is less than 1 minute on a "normal" desktop computer.