Package mml

Package 'mml'; see mml's README.

See: Description

Interface Summary

Interface	Description
HasPdf	Implemented by a (fully parameterised) Model class that has a probability density function HasPdf.pdf(la.la.Value), and a HasPdf.nlPdf(la.la.Value).

Class Summary

Class	Description
Adaptive	The UnParameterised Adapative Model of Bounded Discrete data; also see M.
BestOf	The UnParameterised BestOf Model – choose the best one out of a given Tuple of alternative UnParameterised Models.
BetaUPM	(INCOMPLETE, no Estimator yet) The Β (Beta) Model (probability distribution).
ByPdf	A class that may be useful in defining (UnParameterised) Models over continuous data-spaces such as, but not limited to, Value.Cts, and as defined by a probability density function pdf(d) and nlPdf(d).
Continuous	Continuous may be used to define (UnParameterised) Models over Cts Values by specifying a (-log) probability density function, Continuous.M.nlPdf_x(double), on a Value's 'x()'.
Continuous.Bounded	UnParameterised Bounded Continuous Models, over a range, [lwb, upb], of Cts Values.
Continuous.Uniform	The UnParameterised Uniform Continuous Model on the range [lwb, upb].
CPT	CPT -- Conditional Probability Table -- an (UnParameterised) FunctionModel, from an input datum with a working n(), such as a Value.Discrete or a Tuple of Discretes, to a Model of the output datum.
Dependent	An UnParameterised Dependent Model of data pairs, ⟨id, od⟩, made from an UnParameterised Model, upm, of the input (independent) datum, id, and from an UnParameterised FunctionModel, upfm, of the output (dependent) datum, od, conditional on id.
Direction	UnParameterised Models of Directions, that is of RD-Vectors, of known norm (length, size, radius).
Direction.Uniform	The UnParameterised Uniform Model of Directions in RD.
Dirichlet	The UnParameterised Dirichlet Model (probability distribution) for data from a K-Simplex.
Discretes	The (abstract) sub-class of UnParameterised Model that may be helpful in defining Models of Discrete data-spaces (Values), such as Value.Enum data.
Discretes.Bounded	The class of UnParameterised Models over Bounded Discrete data such as [3, 7], or DNA, say.
Discretes.Uniform	The UnParameterised Uniform Model (distribution) on data in bounds = [lo, hi], pretty much the simplest Discrete Model having only the trivial "statistical parameter." Also see the fully parameterised Discretes.Uniform.M, and Continuous.Uniform.
Estimator	An Estimator estimates (fits) a fully parameterised Model to a data-set, thus solving the inference problem posed by some UnParameterised Model.
ExponentialUPM	The class of UnParameterised (negative-) Exponential Model(s).
FunctionModel	The fully parameterised Function Model (aka regression) with input datum (independent variable), 'id', and output datum (dependent variable), 'od'.
GammaUPM	For most purposes MML.Gamma and GammaUPM.M should be enough, with little need to call upon GammaUPM directly.
Geometric0UPM	The UnParameterised Geometric0 Model.
Graphs	The general abstract UnParameterised Model of the structure of Graphs having explicit gType and upmV where the former is the Type (population) of Graphs and the latter is the Model of the number of Vertices, |V|.
Graphs.GERadaptive	Special case of the Graphs.IndependentEdges Model where upmE() is Adaptive over [0, 1].
Graphs.GERfixed	Common special case of the Graphs.IndependentEdges Model where upmE() is MultiState over [0, 1].
Graphs.IndependentEdges	Extends Graphs to Models where there is also an explicit Model upmE on the existence of Edges.
Graphs.Motifs	UnParameterised Models based on the notion of frequent sub-graphs (motifs, patterns) and that can calculate the message-length contribution of a set of motifs.
Graphs.Skewed	An UnParameterised Model of Graphs where the degree distribution is skewed, few Vertices having high degree and many having low degree.
HeavyTail	Experimental, may change or disappear: Heavy-tailed continuous probability distribution(s).
HeavyTail.Over_x1	A heavy-tailed, log-symmetric, everywhere differentiable, probability distribution for continuous (real-valued) X>0.
Independent	A Tuple of UnParameterised Models makes an UnParameterised Model of Tuples, i.e., of Multivariate data, and a Tuple of statistical parameters make the statistical parameter of the Model, etc..
Int1	A simple model of positive integers, >0, where pr(n) = 1/(n(n+1)).
Intervals	Intervals, the UnParameterised FunctionModel.
KnownClass	See Known Model.
LaplaceUPM	The UnParameterised Laplace.
Linear1	The class of UnParameterised Function Models (regressions) from a Continuous input datum (independent variable), 'x', to a Continuous output datum (dependent variable), 'y', R→R, y=ax+b+N(0,σ).
LinearD	The UnParameterised LinearD function-model; also see the fully parameterised LinearD function-model.
LogStar0UPM	Model MML.logStar0 should be enough for most purposes, but here are the classes, UnParameterised (LogStar0UPM) and fully (trivially) parameterised (M).
Markov	A Markov Model of a given order — the UnParameterised Series Model.
Missing	The UnParameterised Missing Model of "missing data" – data that may be known (present) or may be missing (absent, unknown).
Mixture	An UnParameterised Mixture Model capable of producing a fully parameterised Mixture Model; in this case, the problem-defining parameter is a UPModel, Mixture.upm, and the statistical-parameters are the weights, plus the statParams, of the clusters — the classes in the Snob sense of "class".
MML	The class of Minimum Message Length [MML] tools for statistical and inductive inference, and machine learning.
Model	The abstract class of fully parameterised statistical Models.
Model.Defaults	A subclass of Model that sets default methods stats(ds,lo,hi), stats(add,ss0,ss1) and nlLH(ss), even if they are slow.
MotifA	'MotifA', the UnParameterised adaptive Motif Model of' Graphs.
MotifD	The UnParameterised 'MotifD' Model of Graphs.
Multinomial	An UnParameterised FunctionModel, given a number of categories 'k', from a number of trials 'n' to a Vector of 'k' frequencies that sum to 'n'.
MultiState	The UnParameterised MultiState Model (MultiState distribution) on data in bounds = [lo, hi], capable of estimating a fully parameterised MultiState.M Model.
Multivariate	The (abstract) class of UnParameterised Models over Multivariate data (Tuples).
NaiveBayes	The UnParameterised NaiveBayes FunctionModel; the fully parameterised FunctionModel is NaiveBayes.M.
NearInverse	The UnParameterised NearInverse Model of positive reals, (0, ∞).
NormalMu	An UnParameterised Normal Model for cases where μ is the problem-defining parameter (common knowledge, given) and σ alone is the statistical parameter of the fully parameterised NormalMu.M to be estimated.
NormalUPM	The class of UnParameterised Normal Models (Gaussian distributions).
Permutation	The abstract class of UnParameterised Models over Permutations {0, ..., N-1}.
Permutation.Uniform	The UnParameterised Uniform Model of Permutations.
Poisson0UPM	The UnParameterised Poisson0 Model.
R_D	R_D, the (abstract) class of UnParameterised Models of D-dimensional Vectors of Cts, that is of RD.
R_D.Forest	A (UnParameterised) model in which a column may be linearly dependent on a single parent column.
R_D.ForestSearch	A generalisation of R_D.Forest that is able to estimate the model's structure – the child-parent relations – as well as the parameters of the component distributions.
R_D.Independent	A Vector of Models is a Model of Vectors RD.
R_D.NrmDir	An UnParameterised Model of Vectors in RD made from UnParameterised Models of norms (lengths) and of Directions.
README	About package 'mml' — tools for Minimum Message Length inference.
Sequences	The UnParameterised Sequences Model; the fully parameterised Model is Sequences.M.
Sequences.K	Sequence.K uses the same Model, Sequences.K.eltUPM (parameterised), for every element of every Sequence datum.
Sequences.LtoR	The UnParameterised Left-to-Right Model of Sequences.
SeriesModel	Fully parameterised (Time-) Series Models of data Series (Vectors).
SeriesModel.Analysis	An Analysis of a Scannable Value (Series).
Simplex	For Models of data s : K_Simplex; a standard K_Simplex is the space of (K+1)-Vectors, {[s0, ..., sK]} (NB. 0..K) where 0 ≤ si ≤ 1 and ∑si = 1.
Simplex.Uniform	The UnParameterised Uniform Model over a K_Simplex data-space (Mdl is fully parameterised).
Test	Test.main(java.lang.String[]) runs a few simple tests on mml.
Tree	The class of UnParameterised (Decision | Classification | Regression)- Tree FunctionModels.
Tree.Param	The root class for the statistical parameter of a fully parameterised Tree FunctionModel.
Tree.Param.DFork	The statistical parameter of a Tree.DFork, i.e., an input-column number, and a Vector of parameters for sub-Trees.
Tree.Param.Fork	The superclass of Tree.Param.DFork and Tree.Param.OFork.
Tree.Param.Leaf	The statistical parameter of a Tree.Leaf, i.e., a statistical parameter to make a Tree.Leaf from Tree.leafUPM.
Tree.Param.OFork	The statistical parameter of a Tree.OFork, i.e., an input-column number, a splitting Value (< v.
UPFunctionModel	UnParameterised (abstract) Function Model (aka regression) with input datum (independent variable), 'id', and output datum (dependent variable), 'od'.
UPFunctionModel.K	A very simple UnParameterised Function Model that uses a single Model for every output datum, od, regardless of the input datum, id.
UPModel	UPModel, the abstract class of UnParameterised statistical Models.
UPSeriesModel	The (abstract) class of UnParameterised Models of Series.
UPSeriesModel.K	A very simple, UnParameterised SeriesModel that uses a given UnParameterised Model of length ≥ 0, and a given UnParameterised Model of (every) element.
UPSeriesModel.Length	A subclass of UPSeriesModels where Length.M has an explicit model of lengths (as opposed to a terminating Value/symbol).
vMF	The UnParameterised von Mises - Fisher Model of Directions in RD.
WallaceInt0UPM	Model WallaceInt0 should be enough for most purposes but here are the classes, UnParameterised (WallaceInt0UPM) and fully (trivially) parameterised (M).

Package mml Description

Package 'mml'; see mml's README.