CPU and GPU-accelerated Machine Learning Library
Jupyter Notebook Scala Cuda C++ C Shell Other


BIDMach is a very fast machine learning library. Check the latest benchmarks

The github distribution contains source code only. To get the libraries for your platform, run ./getdevlibs.sh from this directory. Then you can run bidmach with ./bidmach. Also, you can run the iScala notebook using ./bidmach notebook.

Update 01/30/2017: ./getdevlibs.sh is obsolete, so please see the installation instructions here using maven. If you modifythe code for your own use, the command mvn package will compile the code.

You can build the Java/Scala main jar with sbt (included). There are build scripts for both Scala 2.10 and 2.11. Copy the appropriate one into build.sbt, and then do ./sbt package. You can then run bidmach with ./bidmach (you still need to download the libraries some of which are native).

You can also download an executable bundle from here. You will need the libs from there in order to build from a git branch. We use a lot of native code which isn't all available from repos, and you will save a lot of time and headaches by grabbing compiled versions.

The main project page is here.

Documentation is here in the wiki

New BIDMach has a discussion group on Google Groups.

BIDMach is a sister project of BIDMat, a matrix library, which is also on github