Dimension reduction for linear separation with curvilinear distances

Jonathan Winkley, Ping Jiang, Alamgir Hossain

    Research output: Contribution to conferencePaperpeer-review

    2 Citations (Scopus)


    Any high dimensional data in its original raw form may contain obviously classifiable clusters which are difficult to identify given the high-dimension representation. In reducing the dimensions it may be possible to perform a simple classification technique to extract this cluster information whilst retaining the overall topology of the data set. The supervised method presented here takes a high dimension data set consisting of multiple clusters and employs curvilinear distance as a relation between points, projecting in a lower dimension according to this relationship. This representation allows for linear separation of the non-separable high dimensional cluster data and the classification to a cluster of any successive unseen data point extracted from the same higher dimension.
    Original languageEnglish
    Publication statusPublished - 2011
    Event17th International Conference on Soft Computing - Brno, Czech Republic
    Duration: 1 Jan 2011 → …


    Conference17th International Conference on Soft Computing
    Period1/01/11 → …


    Dive into the research topics of 'Dimension reduction for linear separation with curvilinear distances'. Together they form a unique fingerprint.

    Cite this