Class PINN<O extends NumberVector>

  • Type Parameters:
    O - Object type
    All Implemented Interfaces:
    IndexFactory<O>

    @Title("PINN: Projection Indexed Nearest Neighbors")
    @Reference(title="Finding local anomalies in very high dimensional space",
               authors="T. de Vries, S. Chawla, M. E. Houle",
               booktitle="Proc. IEEE 10th International Conference on Data Mining (ICDM)",
               url="https://doi.org/10.1109/ICDM.2010.151",
               bibkey="DBLP:conf/icdm/VriesCH10")
    public class PINN<O extends NumberVector>
    extends ProjectedIndex.Factory<O,​O>
    Projection-Indexed nearest-neighbors (PINN) is an index to retrieve the nearest neighbors in high dimensional spaces by using a random projection based index.

    Reference:

    Finding local anomalies in very high dimensional space
    T. de Vries, S. Chawla, M. E. Houle
    In: Proc. IEEE 10th International Conference on Data Mining (ICDM)

    Since:
    0.6.0
    Author:
    Erich Schubert
    • Constructor Detail

      • PINN

        public PINN​(IndexFactory<O> inner,
                    int t,
                    double s,
                    double h,
                    RandomFactory random)
        Constructor.
        Parameters:
        inner - Inner index
        t - Target dimensionality
        s - Sparsity
        h - Neighborhood size multiplicator
        random - Random generator factory