Skip to content
Open source Machine Learning library written in Java
Branch: master
Clone or download
Type Name Latest commit message Commit time
Failed to load latest commit information.
core increment version to 0.11.3 Apr 18, 2019
phrase-count-plugin increment version to 0.11.3 Apr 18, 2019
.gitignore url encoding Aug 17, 2017
LICENSE.txt add apache license May 26, 2016 Update Sep 29, 2017


A Java Machine Learning Library

Pyramid is a Java machine learning library which implements many state-of-the-art machine learning algorithms, including

  • Binary and Multi-class classification algorithms:
    • Logistic Regression with L1 regularization (Lasso), L2 regularization (Ridge) and L1+L2 regularization (Elastic-net)
    • Variational Bayesian Logistic Regression
    • Gradient Boosted Trees
    • Naive Bayes
    • Error-Correcting Output Codes (ECOC)
    • Support Vector Machines (SVM)
  • Multi-label classification algorithms:
  • Regression algorithms:
    • Linear Regression with L1 regularization (Lasso), L2 regularization (Ridge) and L1+L2 regularization (Elastic-net)
    • Variational Bayesian Linear Regression
    • Regression Tree
    • Gradient Boosted Trees
  • Learning to rank algorithms:
    • LambdaMART
  • Clustering:

At the moment, not all algorithms are released. We are actively working on tidying up the source files and adding documentations. We will release a few algorithms at a time when they are ready and hope to have all algorthms released soon!


If you just want to use pyramid as a command line tool (which is very simple), all you need is Java 8.

If you are also a Java developer and wish to call Pyramid Java APIs, you will also need Maven.


Pyramid doesn't require any installation effort. All you need is downloading the latest pre-compiled package (with a name like and decompressing it. Now you can move into the created folder and type

./pyramid config/

You will see a welcome message and that means everything is working perfectly.

Windows users please see the notes.

Command Line Usage

All algorithms/functions implemented in Pyramid can be run though a simple command, with the following syntax:

./pyramid <properties_file>


./pyramid config/


./pyramid config/

pyramid is a launcher script and <properties_file> is a file specifying the name of the algorithm and all necessary parameters, such as the input data, output folder, and learning algorithm hyper parameters. The <properties_file> can be specified by either an absolute or a relative path.

To run different algorithms, you just need to invoke the program with different properties files. The list of available algorithms and their corresponding properties file templates can be found in the Wiki.

Building from Source

If you are a Java developer who prefer working with the source code or want to contribute to the Pyramid package:

Pyramid uses Maven for its build system.

To compile and package the project from the source code, simply run the mvn clean package -DskipTests command in the cloned directory. The compressed package will be created under the core/target/releases directory.


We welcome your feedback on the package. To ask questions, request new features or report bugs, please contact Cheng Li via

Answers to some commonly asked questions can be found in FAQ.

You can’t perform that action at this time.