Direction numbers obtained using the search criterion D(6) up to the dimension 21201. This is the recommended choice by the authors.

Original data can be found at https://web.maths.unsw.edu.au/~fkuo/sobol/. For additional details on the quantities involved, see [1].

The C-code generated from putting the numbers in as literals is obscenely large/inefficient. The data file was thus packaged and save as an .npz data file for fast loading using the following code (this assumes that the file https://web.maths.unsw.edu.au/~fkuo/sobol/new-joe-kuo-6.21201 is present in the working directory):

import pandas as pd import numpy as np

# read in file content with open("./new-joe-kuo-6.21201", "r") as f: lines = f.readlines()

rows = []

# parse data from file line by line for l in lines[1:]: nums = [int(n) for n in l.replace("

# read in as dataframe, explicitly use zero values df = pd.DataFrame(rows).fillna(0).astype(int)

# perform conversion df["poly"] = 2 * df["a"] + 2 ** df["s"] + 1

# ensure columns are properly ordered vs = df[[f"v{i}" for i in range(18)]].values

# add the degenerate d=1 column (not included in the data file) vs = np.vstack([vs[0][np.newaxis, :], vs]) poly = np.concatenate([[1], df["poly"].values])

# save as compressed .npz file to minimize size of distribution np.savez_compressed("./_sobol_direction_numbers", vinit=vs, poly=poly)

Load direction numbers into two arrays.