Technology & Services: How is the KDE Different?

Most multivariate analyses attempt to reduce complexity of the data streams.  They do so in various ways, such as by:

• filtering the data for peaks
• removing values below an arbitrary threshold
• binning data to reduce the number of data points
• reducing the dimensionality using methods  such as Principal Component Analysis (PCA)

Most also employ linear mathematics, relying on standard statistics such as the mean and standard deviation for feature selection, and standard linear techniques, such as PCA or linear discriminant analysis, for classification.

But the real world is better described by complex geometry with high dimensionality.  Therefore, Correlogic’s technology embraces the complexity and does not seek to alter it with preconceived filters or dimensional reduction.  Instead, Correlogic’s algorithms search for an optimal solution in the complete data space available, using efficient clustering techniques.  While standard statistical analyses have a place, it is at the end of the process -- as a tool to judge the robustness of the solution -- not at the start, where alteration of the data can alter the solution itself.

Correlogic also uses a highly optimized genetic algorithm for feature selection as well as cluster definition to expedite the search for highly discriminatory and robust models.

Together, these differences provide strong analytical advantages to the Correlogic technology.