MLpp

C++ library with efficient implementations of selected machine learning algorithms.

(C) 2020-21 Roman Werpachowski.

Project webpage: https://romanwerpachowski.github.io/ML/

Provided algorithms

Clustering

We provide the Gaussian E-M and K-means algorithms. Initial means can be initialised in three different ways:

• Forgy and Random partition
• K++

Implemented in ml::Clustering namespace and ml::EM class.

Decision trees

Supported decision trees:

• multinomial classification
• multivariate regression with a scalar dependent variable

With and without cost-complexity pruning.

Implemented in ml::DecisionTrees namespace.

Linear regression

Only Ordinary Least Squares for now:

• univariate with and without intercept
• multivariate
• recursive multivariate
• ridge regression
• PRESS statistic

Implemented in ml::LinearRegression namespace.

Logistic regression

Conjugate gradient logistic regression algorithm. See class ml::ConjugateGradientLogisticRegression.

Cross-validation

Methods:

• k-fold
• leave-one-out

Implemented in ml::Crossvalidation namespace.

Statistics

A set of standard statistical functions used by other modules. Implemented in ml::Statistics namespace.