mml

Class FunctionModel

java.lang.Object

la.la.Value

mml.Model

mml.FunctionModel

All Implemented Interfaces:

java.lang.Comparable<Value>

Direct Known Subclasses:

UPFunctionModel.M

public abstract class FunctionModel
extends Model

The fully parameterised Function Model (aka regression) with input datum (independent variable), 'id', and output datum (dependent variable), 'od'. Called 'Function Models' because they are like "noisy", statistical functions, id → od. Most instances of FunctionModel are likely to be instances of UPFunctionModel.M. The main action is in and around condModel(id). Also see Dependent. For a simple but useful example of a Function Model see CPT (conditional probability table). Note that 'id' and/or 'od' may be multivariate. (And, a FunctionModel is a Model, not a Function.-) Also see Model and SeriesModel.

Nested Class Summary

Nested classes/interfaces inherited from class mml.Model

Model.Defaults, Model.Transform

Nested classes/interfaces inherited from class la.la.Value

Value.Atomic, Value.Bool, Value.Char, Value.Chars, Value.Cts, Value.Defer, Value.Discrete, Value.Enum, Value.Inc_Or, Value.Int, Value.Lambda, Value.List, Value.Maybe, Value.Option, Value.Real, Value.Scannable, Value.Structured, Value.Triv, Value.Tuple

Field Summary

Fields inherited from class mml.Model

msg1, msg2, sp

Fields inherited from class la.la.Value

C, comparator, CR, E, eight, ffalse, five, fiveR, four, fourR, half, negCR, negOne, negOneR, negTenR, nilVal, nine, None, one, oneR, PI, PIby2, point01, point1, point5, point9, quarter, seven, six, ten, tenR, three, threeR, triv, ttrue, two, twoR, zero, zeroR

Constructor Summary

Constructors

Constructor and Description
FunctionModel(double msg1, double msg2, Value sp) Construct a fully parameterised FunctionModel with 2-part message length msg1+msg2, and statistical parameter(s), sp.

Method Summary

Modifier and Type	Method and Description
FunctionModel	asGiven(double msg2) Calls asGiven(0, msg2).
FunctionModel	asGiven(double msg1, double msg2) Return a clone of 'this' FunctionModel but with msg1 and msg2 as specified.
UPFunctionModel	asUPModel() Treat 'this' fully parameterised FunctionModel as an UnParameterised one.
abstract Model	condModel(Value id) Return the Model for output datum, od, conditional upon the given input datum, id.
double	condNl2Pr(Value id, Value od) Base 2 (bits) version of condNlPr.
double	condNlPr(Value id, Value od) Return the negative log probability of the output datum, od, conditional upon the input datum, id.
double	condPr(Value id, Value od) Return the probability of the output datum, od, conditional upon the input datum, id.
double	nlPr(Value iod) Given iod = ⟨id, od⟩, return -log pr(od|id).
double	pr(Value iod) Given iod = ⟨id, od⟩, return pr(od|id).
Value	random() random() is inappropriate and throws an Exception but see random(id).
Value	random(Value id) condModel(id).random().

Methods inherited from class mml.Model

asEstimator, m1m2sp, msg, msg1, msg1bits, msg2, msg2bits, msgBits, nl2LH, nl2Pr, nlLH, random, randomSeries, statParams, stats, stats, stats, stats, sumNlPr, toString, transform, type, zeroTriv

Methods inherited from class la.la.Value

AoM, apply, bOp, compareTo, cons, cts, elt, errMsg, error, force, isTuple, just, main, n, nElts, nlAoM, pair, print, quad, real, RTE, triple, tuple, UOE, uOp, x

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

Constructor Detail

FunctionModel

public FunctionModel(double msg1,
                     double msg2,
                     Value sp)

Construct a fully parameterised FunctionModel with 2-part message length msg1+msg2, and statistical parameter(s), sp.

Method Detail

condModel

public abstract Model condModel(Value id)

Return the Model for output datum, od, conditional upon the given input datum, id. condModel(id) is the key thing to define for a new FunctionModel. Also see condNlPr(id,od) and condPr(id,od).

nlPr

public double nlPr(Value iod)

Given iod = ⟨id, od⟩, return -log pr(od|id). The input datum, id, is common knowledge and thus "free" – everyone knows id particularly the receiver (MML). The output datum, od, is conditional upon id. (Also see Dependent.)

Specified by:

nlPr in class Model

pr

public double pr(Value iod)

Given iod = ⟨id, od⟩, return pr(od|id). The input datum, id, is common knowledge and thus "free" – everyone knows id particularly the receiver (MML). The output datum, od, is conditional upon id. (Also see Dependent.)

Overrides:

pr in class Model

condNlPr

public double condNlPr(Value id,
                       Value od)

Return the negative log probability of the output datum, od, conditional upon the input datum, id. This default calls condModel(id).nlPr(od) but a sub-class might know a short cut?

condNl2Pr

public double condNl2Pr(Value id,
                        Value od)

Base 2 (bits) version of condNlPr.

condPr

public double condPr(Value id,
                     Value od)

Return the probability of the output datum, od, conditional upon the input datum, id. Note the order of the parameters. This default calls condModel(id).pr(od) but a sub-class might know a short cut?

asGiven

public FunctionModel asGiven(double msg2)

Calls asGiven(0, msg2).

Overrides:

asGiven in class Model

asGiven

public FunctionModel asGiven(double msg1,
                             double msg2)

Return a clone of 'this' FunctionModel but with msg1 and msg2 as specified. Also see Model.asGiven(,).

Overrides:

asGiven in class Model

asUPModel

public UPFunctionModel asUPModel()

Treat 'this' fully parameterised FunctionModel as an UnParameterised one. Also see Model.asUPModel().

Overrides:

asUPModel in class Model

random

public Value random()

random() is inappropriate and throws an Exception but see random(id). Note, that condModel(id) might implement random(), we hope.

Overrides:

random in class Model

random

public Value random(Value id)

condModel(id).random(). Given an input datum 'id', generate a random output datum. Obviously requires condModel(id) to implement random().