MADlib® is an open-source library for scalable in-database analytics.

It provides data-parallel implementations of mathematical, statistical and machine learning methods for structured and unstructured data.

Quick Start Guides

Get going with a minimum of fuss. 

• Installation Guide
• Quick Start Guide for Users
• Quick Start Guide for Developers

General Information

Learn about MADlib.

• MADlib website 
• Module and algorithm documentation
• FAQ

Developer Documentation

Contribute to the project.

• Source code repo
• Contribution Guidelines
• Documentation Guide (Doxygen)
• Ideas for contribution 

Architecture

See how the pieces fit together. 

• Architecture
• Module Anatomy

Release Notes

See what has been released.

• Last release
• Historical release notes for releases prior to move to ASF. 

Third Party Components

MADlib incorporates material from the following third-party components:

1. argparse 1.2.1 "provides an easy, declarative interface for creating command line tools"
2. Boost 1.47.0 (or newer) "provides peer-reviewed portable C++ source libraries"
3. doxypy 0.4.2 "is an input filter for Doxygen"
4. Eigen 3.2.2 "is a C++ template library for linear algebra"
5. PyYAML 3.10 "is a YAML parser and emitter for Python"
6. PyXB 1.2.4 "is a Python library for XML Schema Bindings"

Licensing

License information regarding MADlib and included third-party libraries can be found inside the license directory.

Papers

• MAD Skills : New Analysis Practices for Big Data (VLDB 2009)
• Hybrid In-Database Inference for Declarative Information Extraction (SIGMOD 2011)
• Towards a Unified Architecture for In-Database Analytics (SIGMOD 2012)
• The MADlib Analytics Library or MAD Skills, the SQL (VLDB 2012)

Related Software

• PivotalR - lets the user run the functions of the open-source big-data machine learning package MADlib directly from R.

• PyMADlib  - a nascent Python wrapper for MADlib, which brings you the power and flexibility of python with the number crunching power of MADlib.