Skip to content

Numo::TinyLinalg is a subset library from Numo::Linalg consisting only of methods used in Machine Learning algorithms.

License

Notifications You must be signed in to change notification settings

yoshoku/numo-tiny_linalg

Numo::TinyLinalg

Gem Version Build Status BSD 3-Clause License Documentation

Numo::TinyLinalg is a subset library from Numo::Linalg consisting only of methods used in Machine Learning algorithms. The functions Numo::TinyLinalg supports are dot, det, eigh, inv, pinv, qr, solve, cholesky, cho_solve and svd.

Note that the version numbering rule of Numo::TinyLinalg is not compatible with that of Numo::Linalg.

Installation

Unlike Numo::Linalg, Numo::TinyLinalg only supports OpenBLAS as a backend library for BLAS and LAPACK.

Install the OpenBlas.

macOS:

$ brew install openblas

Ubuntu:

$ sudo apt-get install libopenblas-dev liblapacke-dev

Install the gem and add to the application's Gemfile by executing.

macOS:

$ bundle config --local build.numo-tiny_linalg "--with-opt-dir=/opt/homebrew/Cellar/openblas/0.3.23/"
$ bundle add numo-tiny_linalg

Ubuntu:

$ bundle add numo-tiny_linalg

If bundler is not being used to manage dependencies, install the gem by executing.

macOS:

$ gem install numo-tiny_linalg -- --with-opt-dir=/opt/homebrew/Cellar/openblas/0.3.23/

Ubuntu:

$ gem install numo-tiny_linalg

Usage

An example of singular value decomposition.

require 'numo/tiny_linalg'

Numo::Linalg = Numo::TinyLinalg unless defined?(Numo::Linalg)

x = Numo::DFloat.new(5, 2).rand.dot(Numo::DFloat.new(2, 3).rand)
# =>
# Numo::DFloat#shape=[5,3]
# [[0.104945, 0.0284236, 0.117406],
#  [0.862634, 0.210945, 0.922135],
#  [0.324507, 0.0752655, 0.339158],
#  [0.67085, 0.102594, 0.600882],
#  [0.404631, 0.116868, 0.46644]]

s, u, vt = Numo::Linalg.svd(x, job: 'S')

z = u.dot(s.diag).dot(vt)
# =>
# Numo::DFloat#shape=[5,3]
# [[0.104945, 0.0284236, 0.117406],
#  [0.862634, 0.210945, 0.922135],
#  [0.324507, 0.0752655, 0.339158],
#  [0.67085, 0.102594, 0.600882],
#  [0.404631, 0.116868, 0.46644]]

puts (x - z).abs.max
# => 4.440892098500626e-16

Contributing

Bug reports and pull requests are welcome on GitHub at https://github.com/yoshoku/numo-tiny_linalg. This project is intended to be a safe, welcoming space for collaboration, and contributors are expected to adhere to the code of conduct.

Code of Conduct

Everyone interacting in the Numo::TinyLinalg project's codebases, issue trackers, chat rooms and mailing lists is expected to follow the code of conduct.

License

The gem is available as open source under the terms of the BSD-3-Clause License.

About

Numo::TinyLinalg is a subset library from Numo::Linalg consisting only of methods used in Machine Learning algorithms.

Topics

Resources

License

Code of conduct

Stars

Watchers

Forks

Packages

No packages published

Languages