A Concept Learning Tool Based On Calculating Version Space Cardinality
This addresses concept learning for imbalanced datasets, but it is incremental as it builds on existing version space methods with computational enhancements.
The paper tackled concept learning on extremely imbalanced datasets where cross-validation is not feasible by proposing VeSC-CoL, which uses version space cardinality as a model quality measure and employs Ordered Binary Decision Diagram and Boolean Satisfiability for computation, achieving accurate learning of the target concept when computational resources are available.
In this paper, we proposed VeSC-CoL (Version Space Cardinality based Concept Learning) to deal with concept learning on extremely imbalanced datasets, especially when cross-validation is not a viable option. VeSC-CoL uses version space cardinality as a measure for model quality to replace cross-validation. Instead of naive enumeration of the version space, Ordered Binary Decision Diagram and Boolean Satisfiability are used to compute the version space. Experiments show that VeSC-CoL can accurately learn the target concept when computational resource is allowed.