ruby-alglib

• ALGLIB for Ruby.
• ALGLIB is a numerical analysis and data processing library. This repository provides bindings primarily for statistics.

Installation

From source code:

git clone https://github.com/kojix2/ruby-alglib
cd ruby-alglib
bundle install
rake ext:alglib # Download ALGLIB Free Edition from https://www.alglib.net/ and extract to ext/alglib.
rake compile
rake install

• Note that the license of ALGLIB Free Edition is GPL or Personal/Academic license.
• If you want to use ALGLIB in a commercial application, you will need to purchase a commercial license.
  • Can I use ALGLIB Free Edition in a commercial application which runs on server?
• You can also use the Commercial Edition of ALGLIB. In that case, you need to replace the files in ext/alglib with the files in the Commercial Edition.
• Windows is not supported.

Usage

Supported Statistical Functions

Special Functions (from alglib_specialfunctions.h)

• gamma_function
• lngamma
• error_function
• error_functionc
• normal_pdf
• normal_cdf
• inv_erf
• inv_normal_cdf
• bivariate_normal_pdf
• bivariate_normal_cdf
• incomplete_beta
• inv_incomplete_beta
• student_t_distribution
• inv_student_t_distribution
• f_distribution
• fc_distribution
• inv_f_distribution
• incomplete_gamma
• incomplete_gamma_c
• inv_incomplete_gamma_c
• chi_square_distribution
• chi_square_c_distribution
• inv_chi_square_distribution
• binomial_distribution
• binomial_c_distribution
• inv_binomial_distribution
• exponential_integralei
• exponential_integralen
• poisson_distribution
• poisson_c_distribution
• inv_poisson_distribution
• beta

Data Analysis Functions (from alglib_dataanalysis.h)

• pca_build_basis

Miscellaneous Functions (from alglib_alglibmisc.h)

• hqrnd_normal
• hqrnd_normalv

1D Array Functions

• sample_moments
• sample_mean
• sample_variance
• sample_skewness
• sample_kurtosis
• sample_adev
• sample_median
• sample_percentile
• cov2
• pearson_corr2
• spearman_corr2

2D Array / Matrix Functions

• cov_matrix, cov_matrix_with_size
• pearson_corr_matrix, pearson_corr_matrix_with_size
• spearman_corr_matrix, spearman_corr_matrix_with_size
• cov_matrix2, cov_matrix2_with_size
• pearson_corr_matrix2, pearson_corr_matrix2_with_size
• spearman_corr_matrix2, spearman_corr_matrix2_with_size
• rank_data, rank_data_with_size
• rank_data_centered, rank_data_centered_with_size

Statistical Tests

• pearson_correlation_significance
• spearman_rank_correlation_significance
• jarque_bera_test
• f_test
• one_sample_variance_test
• wilcoxon_signed_rank_test
• mann_whitney_u_test
• student_test_1
• student_test_2
• unequal_variance_t_test

Example

require 'alglib'

# 1D statistics
arr = [1.0, 2.0, 3.0, 4.0]
mean = Alglib.sample_mean(arr)
variance = Alglib.sample_variance(arr)
moments = Alglib.sample_moments(arr) # => { "mean" => ..., "variance" => ..., ... }

# 2D statistics
mat = [[1, 2], [3, 4], [5, 6]]
cov = Alglib.cov_matrix(mat)
corr = Alglib.pearson_corr_matrix(mat)
ranked = Alglib.rank_data(mat)
centered_ranked = Alglib.rank_data_centered(mat)

# with_size version
cov2 = Alglib.cov_matrix_with_size(mat, 3, 2)

# Statistical test
result = Alglib.pearson_correlation_significance(0.8, 10) # => { "bothtails" => ..., ... }

Testing

• This project includes comprehensive tests for all statistical functions, special functions, data analysis, and miscellaneous utilities.
• All results are compared with R's standard statistical functions (e.g., mean, var, cor, cov, rank, etc.) for correctness.
• Both 1D and 2D array functions, all with_size variants, and special/data analysis/miscellaneous functions are covered by the test suite.
• Rank-related functions are compared as sets due to possible differences in tie-breaking order between ALGLIB and R.
• See test/statistics_test.rb, test/dataanalysis_test.rb, and other test files for details.

Development

Not-To-Do List

• Creating namespaces under Alglib. (e.g. Alglib::Statistics)
• Trying to implement all functions even if you do not plan to use them.
• Adding features that will require frequent maintenance.
• Enforcing the above policies on forked repos.

Rice - C++ binding

We use Rice to create Ruby's C++ binding.

Ruby Gem Release

1. Update the version number in lib/alglib/version.rb.
2. Clean up the extension files:

rake ext:clean_alglib

3. Download the ALGLIB Free Edition:

rake ext:alglib

4. Publish the alglib4 gem:

rake release

Call for committers and owners

This project is looking for a committer and an owner. The original developer, kojix2, is not familiar with the C++ programming language or mathematical calculations. If there is a suitable owner, kojix2 will gladly transfer this project (provided that kojix2 can maintain commit rights).

License

• ruby-alglib itself (the code in this repository) is distributed under the MIT license.
• However, ALGLIB for C++ Free Edition is under the GPL 2+ license.
• The algblib4 gem distributed with the RubyGem server is GPL 2+ licensed because it contains ALGLIB for C++ Free Edition code.
• Please read FAQ carefully before you use ALGLIB to understand what is allowed in the free version.
  • Can I use ALGLIB Free Edition in a commercial application which runs on server?

    You can, but there is at least one reason to buy commercial license. Everything is OK while it is just running on your server. But any action connected with distribution - selling, sublicensing, transferring to someone under nondisclosure agreement - will trigger copyleft clause. So it is safer and more convenient to buy commercial license.

• Theoretically, ruby-alglib should work with the commercial version. That is why it is MIT. But we have not verified this.