data2semantics/mustard

mustard - Machine learning Using Svms To Analyse Rdf Data

mustard's Issues

Use SimpleGraph for kernel implementations

To remove the dependency on the nodes library, all the kernels should be reimplemented using the SimpleGraph class, already available in org.data2semantics.mustard.simplegraph

Test Issue

Just to test issue with Eclipse

Create Parameter object for the kernels

Create a parameter class to use during the creation of a kernel, to avoid the long and incomprehensible list of parameters of the kernels.

Refactor all the RDF wrapping per kernel

All the kernels in the org.data2semantics.mustard.kernels.rdfdata package are essentially wrappers to easily use and RDFDataSet as input.
This wrapping can be generalized into one or two wrapper classes, similarly to the Hub Removal Kernels.The current invidual wrappers can remain, but implemented via this general wrapper.

Code duplication in WL and Walk Count kernels

Initially the Tree/Root/Approx variants of the kernels were not foreseen, which has lead to the current design. As a consequence these kernels now duplicate a considerable amount of code, which should be refactored. Something similar has occured in the implementation of the WL algorithm, the core of the WL algorithm is repeated a number of times among the different implementation, with the essential differences only being the rewrite step.