A toolbox of AI modules written in Swift: Graphs/Trees, Support Vector Machines, Neural Networks, PCA, K-Means, Genetic Algorithms
Swift Other



A toolbox of AI modules written in Swift: Graphs/Trees, Linear Regression, Support Vector Machines, Neural Networks, PCA, KMeans, Genetic Algorithms, MDP, Mixture of Gaussians, Logistic Regression

This framework uses the Accelerate library to speed up computations. Written for Swift 3.0. Earlier versions are Swift 2.2 compatible

SVM ported from the public domain LIBSVM repository See https://www.csie.ntu.edu.tw/~cjlin/libsvm/ for more information

The Metal Neural Network uses the Metal framework for a Neural Network using the GPU. While it works in preliminary testing, more work could be done with this class

Use the XCTest files for examples on how to use the classes

Playgrounds for Linear Regression, SVM, and Neural Networks are available. Now available in both macOS and iOS versions.

New iOS playgrounds - including a four-in-row game using alpha-beta search!


I have started a manual for the framework. It is a work-in-progress, but adds some useful explanation to pieces of the framework. All protocols, structures, and enumerations are well defined. Class descriptions are there, but not class variables and methods.

Classes/Algorithms supported:

    Depth-first search
    Breadth-first search
    Hill-climb search
    Beam Search
    Optimal Path search

Alpha-Beta (game tree)

Genetic Algorithms
    integer/double alleles

Constraint Propogation
    i.e. 3-color map problem

Linear Regression
    arbitrary function in model
    regularization can be used
    convenience constructor for standard polygons
    Least-squares error

Non-Linear Regression

Logistic Regression
    Use any non-linear solution method
    Multi-class capability

Neural Networks
    multiple layers, several non-linearity models
    on-line and batch training
    feed-forward or simple recurrent layers can be mixed in one network
    simple network training using GPU via Apple's Metal
    LSTM network layer implemented - needs more testing
    gradient check routines

Support Vector Machine
    More-than-2 classes classification

    unlabelled data grouping

Principal Component Analysis
    data dimension reduction

Markov Decision Process
    value iteration
    policy iteration
    fitted value iteration for continuous state MDPs - uses any Regression class for fit
            (see my MDPRobot project on github for an example use)
    Monte-Carlo (every-visit, and first-visit)

    Single variable
    Multivariate - with full covariance matrix or diagonal only

Mixture Of Gaussians
    Learn density function of a mixture of gaussians from data
    EM algorithm to converge model with data

    Use to select model or parameters of model
    Simple validation (percentage of data becomes test data)
    N-Fold validation

    Convolution layers
    Pooling layers
    Fully-connected NN layers

    NSView based MLView for displaying regression data, classification data, functions, and classifier areas!
    UIView based MLView for iOS applications, same as NSView based for macOS

Regression Plot Image Classification Plot Image


This framework is made available with the Apache license.


See the contribution document for information on contributing to this framework