We currently do not provide ELKI on Maven. Thus, there currently is no Maven dependency.
ELKI is changing quickly, and we do not provide a stable API.
For example, in the next release NumberVector<? extends Number>
will simplify to just NumberVector
. Getting rid of this generic is nice, but it will break your code. And such changes still happen at every version, which is why we have not released version 1.0 yet.
If possible, use the command line or MiniGUI. This works really well. I use this to distribute my experiments on a cluster of 50 CPU cores. Only for very advanced cases (involving nontrivial custom distance functions), I have to drive ELKI from Java.
As far as I know, ELKI has the only complete OPTICS implementation. There are some partial implementation such as the one in Weka, but our implementation is much more powerful, and much faster (in particular when you enable indexes). Functionality such as the Xi method for cluster extraction is not available in other implementations at all. And the next ELKI version will also have HDBSCAN*, a revisited version of OPTICS.
Update: ELKI can now be used via Maven. Clone the project https://github.com/elki-project/example-elki-project to get started quickly. The ELKI maven dependencies are:
<!-- ELKI core, without visualization -->
<dependency>
<groupId>de.lmu.ifi.dbs.elki</groupId>
<artifactId>elki</artifactId>
<version>0.7.0</version>
</dependency>
<!-- You only need this dependency if you need visualization -->
<dependency>
<groupId>de.lmu.ifi.dbs.elki</groupId>
<artifactId>elki-batik-visualization</artifactId>
<version>0.7.0</version>
</dependency>