Algorithms in Chameleon Statistics
Seventh Sense Software actively takes part in multivariate statistical
research. The company has developed several state-of-the-art statistical
algorithms and has made a number of important mathematical discoveries in pure
statistics. Amongst them are the following:
Development of the world's fastest single-link clustering algorithm for
Euclidean data sets. The algorithm is both storage efficient, coping easily with
enormous data sets, and parallelizable for time-critical applications.
Development of advanced multi-resolution edge detection techniques for
removal of mixed pixels from multi-spectral imagery to enhance results from
density-based hierarchical clustering techniques such as the single-link method.
Discovery of an important one-to-one relationship between the minimal
spanning tree resulting from single-link clustering of a data set and its sample
density distribution, leading to the development of the so-called "bubble
model" density estimator.
Development of novel techniques in sampling theory leading to the development
of improved parameter estimators and hypothesis testing methods for stochastic
(e.g. Poisson) processes.
Development of an exact k-nearest neighbors density estimator based upon the
above which improves upon the standard estimators.
Development of analytically rigorous boosting-analogues for density-based
classifiers such as k-nearest-neighbors and kernel-based techniques.