This Java package aims at making it easy to quickly write readable and efficient code for general scientific computing. It is a toolbox that tries to be coherent, concise, and easy to integrate.
So far, jpack provides
- Linear algebra primitives
- An implementation of CMA-ES, a black-box, derivative-free optimization algorithm . It's a state-of-the-art, a strong benchmark champion, it's used in the industry, and it has strong theoritical foundations. A few variations are provided, so that one can exploits hypothesis on the optimization problem.
- A very bare-bone implementation of feed-forward neural network.
I (the original author) made jpack to help a friend. A few years later, I have been encouraged to release this code. I plan to add tools in jpack as time goes.
An Ant build file is provided to automate jpack build. From the main directory, do
ant
jpack is going to be compiled and packaged as a single .jar file in the dist directory, as jpack.jar.
You can delete all the build files and distribution files by doing
ant clean
Unit testing is done with JUnit 4.12, and is managed with Ant.
To build and run the unit tests, do
ant test
The JUnit .jar files are not included in the source code, you will have to download them and copy them in a directory named vendor, at the root of the source tree.
The following commands show how you can do this
mkdir vendor
cd vendor
wget http://search.maven.org/remotecontent?filepath=junit/junit/4.12/junit-4.12.jar
wget http://search.maven.org/remotecontent?filepath=org/hamcrest/hamcrest-core/1.3/hamcrest-core-1.3.jar
cd ..
- Alexandre Devert - Initial work - marmakoide
This project is licensed under the MIT License - see the LICENSE.md file for details