ELKI Default parameter settings for (clustering) algorithms

Question

I'm working on a research project that involves showing visually the result of many clustering results. I saw that the clustering classes always bring their own parameterizer static classes.

Is it somehow possible to instantiate all elki algorithms with predefined parameter settings?

Thank you so much!

Answer 1

Yes, you can use the Parameterization API to instantiate and get default values for those parameters, that have default values.

https://elki-project.github.io/howto/java_api#ParameterizationAPI

The current development version has a new "fluent" API, called ELKIBuilder .

Most unit tests have been rewritten to use this, eg, AGNESTest

Clustering<?> clustering = new ELKIBuilder<>(CutDendrogramByNumberOfClusters.class) //
        .with(CutDendrogramByNumberOfClusters.Parameterizer.MINCLUSTERS_ID, 3) //
        .with(AbstractAlgorithm.ALGORITHM_ID, AGNES.class) //
        .with(AGNES.Parameterizer.LINKAGE_ID, SingleLinkage.class) //
        .build().run(db);