tangzk / admm

Hadoop Implementation of the ADMM algorithm

ADMM

ADMM is a Hadoop Map Reduce implementation of the Alternating Direction Method of Multipliers (ADMM) algorithm [1].

Details covering the implementation and statistical background are available in our IEEE BigData 2013 paper and our presentation.

ADMM is a generic optimization algorithm. This implementation includes a logistic regression objective function, however it can be extended to use other objective functions.

To checkout the code and run an example in MapReduce standalone mode, execute:

git clone [e-mail]:intentmedia/admm.git
cd admm
gradle jar
hadoop jar build/libs/admm.jar com.intentmedia.admm.AdmmOptimizerDriver \
-inputPath src/test/java/com/intentmedia/admm/files/logreg_features \
-outputPath output

Dependencies

• gradle

Running Tests

gradle check