libagf (Adaptive Gaussian filtering) is a simple and powerful implementation of variable bandwidth kernel estimators for classification, PDF estimation and interpolation. The library incorporates several innovations to produce one of the fastest and most accurate supervised classification algorithms in the world. These include:
· matching kernel width to sample density quickly and accurately via the properties of the exponential function
· restricting calculations to a set of k-nearest-neighbours found in n log k time with a binary tree
· generating a pre-trained model by searching for the class-borders with guaranteed, superlinear convergence
· extrapolating the conditional probabilities to provide solid knowledge of estimate accuracy
What's New in This Release:
· Finished the QUICKSTART file. This contains (or should contain) everything the beginning user needs to get up and running with the package.
· Also added the paper describing the theory of Adaptive Gaussian Filtering.
· Changed the name of the whole repository to libagf, same as project name.
Product's homepage
What's New in This Release: [ read full changelog ]
· Everything except the I/O routines has been templated.
· With the exception of those used in external routines, variable types in the main routines are now controlled with global typedefs, with each class of variable having a different type.
· Different metrics are now only supported in the routines where they make sense: KNN classification and KNN interpolation.
· The functions now require a pointer to the desired metric.
· The nfold routine now supports interpolation.
· Note that this is still not well tested (if at all).
Find us on Google+
