A simple regulatory architecture allows learning the statistical structure of a changing environment

Published: 19 July 2021| Version 2 | DOI: 10.17632/5xngwv5kps.2
Contributors:
Mikhail Tikhonov,
,

Description

Python 3.7.4 simulation code and scripts reproducing all figures in Landmann, Holmes, Tikhonov (2021) "A simple regulatory architecture allows learning the statistical structure of a changing environment". Optional pre-computed simulation data included to speed-up figure plotting; remove or rename any data file to rerun the relevant simulations from scratch. Code by Stefan Landmann.

Files

Steps to reproduce

Each folder corresponds to one figure in the publication. To create a figure simply run the "_create.py" script in the corresponding folder. When the "_data.py" files are deleted, the script will run the "_simulation.py" files which recreate the data. This may take a long time, expecially when executed on a computer with a small number of cores. General comment: * Code uses old notation of the number of regulators S=N_a and the number of the dimensions of the demand N=N_x. All other deviations of the notation from the publication are indicated in the code. Runtime: * using precomputed data (plotting only): <10 sec * recomputing everything from scratch and using same parameters as in the paper: ~ up to 100 hours on a 40-core machine Troubleshooting: * If, after starting the _create.py file, all data files are created but there is no figure, rerun the _create.py file. If this does not help, run the _plot.py file.