Residual-Entropy

This repository contains code used to generate plots and results discussed in the paper "Residual Entropy" (see https://arxiv.org/abs/1907.03888)

The code is well-tested in a Python 2.7+ environment, but should work fine in Python 3. Let me know by opening an Issue if I need to make tweaks to ensure this.

The two main scripts are:

• fitting_white_noise_sinusoids_1d.py
• fitting_white_noise_chebyshev_1d.py

...and the plot_white_noise_1d.py script creates the plots using outputs from these two.

The scripts are simple, and duplicate each other heavily: this was a deliberate choice to make them each as clear to follow as possible. The required libraries are recent versions of numpy and matplotlib.

To reproduce results in the paper:

1. Run $ python fitting_white_noise_sinusoids_1d.py. This will take a few minutes to run and creates a large pickle file that the plotting script will use to make some of the early charts in the paper.

2. Open a text editor and make the following edits to the Setup parameters in fitting_white_noise_sinusoids_1d.py:

   • Set output_filename = wns1d.1e5.pickle
   • Set store_all = False
   • Set Nruns = 100000

   This script is then ready to run all 10^5 simulations, for which it needs to not retain and store all Y samples and corresponding R residuals in order to prevent you running out of memory!

3. Run $ python fitting_white_noise_sinusoids_1d.py again. This now runs the suite of 10^5 simulations used later in the paper, and may take significantly over an hour.

4. Run $ python fitting_white_noise_chebyshev_1d.py. This will take multiple hours, so be patient.

5. Finally run $ python plot_white_noise_1d.py to make the figures used in the paper.