Kernel Methods

KernelMethods.jl is a library that implements and explores Kernel-Based Methods for supervised learning and semi-supervised learning.

To start using KernelMethods.jl just type into an active Julia session

using Pkg
pkg"add https://github.com/sadit/KernelMethods.jl"

using KernelMethods

KernelMethods.jl consists of the following parts

• Scores. It contains several common performance measures, i.e., accuracy, recall, precision, f1, precision_recall.
• CrossValidation. Some methods to perform cross validation, all of them work through callback functions:
  • montecarlo
  • kfolds
• Supervised. It contains methods related to supervised learning
  • NearNeighborClassifier. It defines a KNN classifier
  • optimize!
  • predict
  • predict_proba

Note: user defined distance functions are accepted; several common distances can be found in SimilaritySearch.jl

KernelMethods.jl depends on

• SimilaritySearch.jl

To reach maximum performance, please ensure that Julia has access to the specific instruction set of your CPUs

http://docs.julialang.org/en/latest/devdocs/sysimg/