Libagf is a software that provides a new way of running adaptive or variable-bandwidth kernel-based estimators to classify statistics effectively.
One of the features that make this software stand out is the ability to match kernel width to sample density quickly and precisely by utilizing the properties of the exponential function. Additionally, calculating a set of k-nearest-neighbors, which are found in n log k time with a binary tree, helps to restrict the calculations. This software also generates a pre-trained model by searching for class-borders with guaranteed, superlinear convergence.
Another novel feature of Libagf is the ability to extrapolate the conditional probabilities, which provides solid knowledge of the estimate's accuracy. This feature provides the user with a reliable result and saves time in further analysis.
In the latest version, the QUICKSTART file has been completed, and it contains everything the beginning user needs to get started with the package. The software also includes a paper that describes the theory of Adaptive Gaussian Filtering. Finally, the name of the repository has been changed to libagf, same as project name.
In summary, Libagf is a must-have software for anyone looking for fast and accurate classification algorithms. The available features offer various options that enable the user to get a reliable result in a short amount of time.
Version 0.9: N/A