Skip to content

Latest commit

 

History

History
72 lines (56 loc) · 3.13 KB

README.md

File metadata and controls

72 lines (56 loc) · 3.13 KB

sparsal - Sparse Approximation Library

Build Status codecov Chat MIT licensed Maven Central

Overview

The Sparse Approximation Library aims to provide easy-to-read, modern and maintainable sparse approximation algorithms written in a functional style.
At the moment only the matching pursuit algorithm is implemented. Future versions will increment the number of algorithms.
Check this article for some basic theoretical background.

Using sparsal

Builds are available for scala 2.10.x, 2.11.x and 2.12.2.

SBT

Simply add sparsal's dependency to your build.sbt if you don't know which version of scala you are using and sbt will automatically detect and search for the appropriate package:

libraryDependencies += "com.github.underscorenico" %% "sparsal" % "0.2.0"

Or either specify the desired target version:

libraryDependencies += "com.github.underscorenico" % "sparsal_2.12" % "0.2.0"

Maven

If you are using Maven, simply add sparsal to your dependencies to your pom.xml:

<dependency>
    <groupId>com.github.underscorenico</groupId>
    <artifactId>sparsal_2.12</artifactId>
    <version>0.2.0</version>
</dependency>

Try it

First initialize your MatchingPursuit (or Orthogonal Matching Pursuit) object with an input and a defined dictionary:

val input = Array.fill(10)(5)

val dictionary = Gabor(input.length)

val mp =  MatchingPursuit1D(input, dictionary)

Then recover the approximation (after the algorithm converges to a certain provided accuracy measure):

    val result: (List[(Double, Int)], Seq[Double]) = mp.run(SNR(20.0))

The result will contain a list of selected atoms from the dictionary. The first element of the touple (type Double) will contain the inner product between the input at Iteration n-1 and the selected atom.
The second element contains the index of the atom from the dictionary.

Contributing

Everyone is welcome to contribute, either by adding features, solving bugs or helping with documentation.
Sparsal embraces the open code of conduct from the TODO group, therefore all of its channels should respect its guidelines.
That being said, we are a community and we should not need guidelines for our conduct but leave it to our common sense instead.