Marek Orliński, MSc, Department of Informatics, NCU: “Fast Encoding Length”
Prototype methods allow to select dataset instances, which are crucial for classification (e.g. kNN). This enables us to reduce dataset, while keeping classification accuracy.
Encoding Length is one of prototype methods with O(n^3) complexity. We propose Locality Sensitive Forest based Encoding Length, which has O(nlogn) complexity, while maintaining the same accuracy.