org.tribuo.regression

Class Regressor

java.lang.Object

org.tribuo.regression.Regressor

All Implemented Interfaces:

Serializable, Iterable<Regressor.DimensionTuple>, Output<Regressor>

Direct Known Subclasses:

Regressor.DimensionTuple

public class Regressor
extends Object
implements Output<Regressor>, Iterable<Regressor.DimensionTuple>

An Output for n-dimensional real valued regression.

In addition to the regressed values, it may optionally contain variances. Otherwise the variances are set to Double.NaN.

Within a DataSource or Dataset each Regressor must contain the same set of named dimensions. The dimensions stored in a Regressor are sorted by the natural ordering of their names (i.e., using the String comparator). This allows the use of direct indexing into the elements.

Note fullEquals(org.tribuo.regression.Regressor) compares the dimensions, the regressed values and the variances. However unlike Double.equals(java.lang.Object), if the two variances being compared are set to the sentinel value of Double.NaN, then they are considered equal.

See Also:

Serialized Form

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	Regressor.DimensionTuple A Regressor which contains a single dimension, used internally when the model implementation doesn't natively support multi-dimensional regression outputs.

Field Summary

Fields

Modifier and Type	Field and Description
static String	DEFAULT_NAME Default name used for dimensions which are unnamed when parsed from Strings.
static double	TOLERANCE The tolerance value for determining if two regressed values are equal.

Constructor Summary

Constructors

Constructor and Description
Regressor(Regressor.DimensionTuple[] dimensions) Constructs a regressor from the supplied dimension tuples.
Regressor(String[] names, double[] values) Constructs a regressor from the supplied named values.
Regressor(String[] names, double[] values, double[] variances) Constructs a regressor from the supplied named values.
Regressor(String name, double value) Constructs a regressor containing a single dimension, using Double.NaN as the variance.
Regressor(String name, double value, double variance) Constructs a regressor containing a single dimension.

Method Summary

Modifier and Type	Method and Description
Regressor	copy() Deep copy of the output up to it's immutable state.
static Regressor	createFromPairList(List<com.oracle.labs.mlrg.olcut.util.Pair<String,Double>> dimensions) Creates a Regressor from a list of dimension tuples.
boolean	equals(Object o) Regressors are equal if they have the same number of dimensions and equal dimension names.
static String[]	extractNames(OutputInfo<Regressor> info) Extracts the names from the supplied Regressor domain in their canonical order.
boolean	fullEquals(Regressor other) Compares other to this output.
Regressor.DimensionTuple	getDimension(int idx) Returns a dimension tuple for the requested dimension index.
Optional<Regressor.DimensionTuple>	getDimension(String name) Returns a dimension tuple for the requested dimension, or optional empty if it's not valid.
String	getDimensionNamesString() Returns a comma separated list of the dimension names.
String	getDimensionNamesString(char separator) Returns a delimiter separated list of the dimension names.
String[]	getNames() The names of the dimensions.
String	getSerializableForm(boolean includeConfidence) Generates a String suitable for writing to a csv or json file.
double[]	getValues() Returns the regression values.
double[]	getVariances() The variances of the regressed values, if known.
int	hashCode() Regressor's hashcode is based on the hash of the dimension names.
Iterator<Regressor.DimensionTuple>	iterator()
static com.oracle.labs.mlrg.olcut.util.Pair<String,Double>	parseElement(int idx, String s) Parses a string of the form:
static Regressor	parseString(String s) Parses a string of the form:
static Regressor	parseString(String s, char splitChar) Parses a string of the form:
int	size() Returns the number of dimensions in this regressor.
String	toString()

Methods inherited from class java.lang.Object

clone, finalize, getClass, notify, notifyAll, wait, wait, wait

Methods inherited from interface java.lang.Iterable

forEach, spliterator

Field Detail

TOLERANCE

public static final double TOLERANCE

The tolerance value for determining if two regressed values are equal.

See Also:

Constant Field Values

DEFAULT_NAME

public static final String DEFAULT_NAME

Default name used for dimensions which are unnamed when parsed from Strings.

See Also:

Constant Field Values

Constructor Detail

Regressor

public Regressor(String[] names,
                 double[] values,
                 double[] variances)

Constructs a regressor from the supplied named values. Throws IllegalArgumentException if the arrays are not all the same size.

Parameters:

names - The names of the dimensions.

values - The values of the dimensions.

variances - The variances of the specified values.

Regressor

public Regressor(String[] names,
                 double[] values)

Constructs a regressor from the supplied named values. Uses Double.NaN as the variances.

Parameters:

names - The names of the dimensions.

values - The values of the dimensions.

Regressor

public Regressor(Regressor.DimensionTuple[] dimensions)

Constructs a regressor from the supplied dimension tuples.

Parameters:

dimensions - The named values to use.

Regressor

public Regressor(String name,
                 double value)

Constructs a regressor containing a single dimension, using Double.NaN as the variance.

Parameters:

name - The name of the dimension.

value - The value of the dimension.

Regressor

public Regressor(String name,
                 double value,
                 double variance)

Constructs a regressor containing a single dimension.

Parameters:

name - The name of the dimension.

value - The value of the dimension.

variance - The variance of this value.

Method Detail

size

public int size()

Returns the number of dimensions in this regressor.

Returns:

The number of dimensions.

getNames

public String[] getNames()

The names of the dimensions. Always sorted by their natural ordering.

Returns:

The names of the dimensions.

getValues

public double[] getValues()

Returns the regression values.

The index corresponds to the index of the name.

In a single dimensional regression this is a single element array.

Returns:

The regression values.

getVariances

public double[] getVariances()

The variances of the regressed values, if known. If the variances are unknown (or not set by the model) they are filled with Double.NaN.

The index corresponds to the index of the name.

In a single dimensional regression this is a single element array.

Returns:

The variance of the regressed values.

toString

public String toString()

Overrides:

toString in class Object

getDimension

public Optional<Regressor.DimensionTuple> getDimension(String name)

Returns a dimension tuple for the requested dimension, or optional empty if it's not valid.

Parameters:

name - The dimension name.

Returns:

A tuple representing that dimension.

getDimension

public Regressor.DimensionTuple getDimension(int idx)

Returns a dimension tuple for the requested dimension index.

Parameters:

idx - The dimension index.

Returns:

A tuple representing that dimension.

Throws:

IndexOutOfBoundsException - if the index is outside the range.

iterator

public Iterator<Regressor.DimensionTuple> iterator()

Specified by:

iterator in interface Iterable<Regressor.DimensionTuple>

copy

public Regressor copy()

Description copied from interface: Output

Deep copy of the output up to it's immutable state.

Specified by:

copy in interface Output<Regressor>

Returns:

A copy of the output.

getSerializableForm

public String getSerializableForm(boolean includeConfidence)

Description copied from interface: Output

Generates a String suitable for writing to a csv or json file.

Specified by:

getSerializableForm in interface Output<Regressor>

Parameters:

includeConfidence - Include whatever confidence score the label contains, if known.

Returns:

A String representation of this Output.

fullEquals

public boolean fullEquals(Regressor other)

Description copied from interface: Output

Compares other to this output. Uses all score values and the strings.

Specified by:

fullEquals in interface Output<Regressor>

Parameters:

other - Another output instance.

Returns:

True if the other instance has value equality to this instance. False otherwise.

equals

public boolean equals(Object o)

Regressors are equal if they have the same number of dimensions and equal dimension names.

Overrides:

equals in class Object

Parameters:

o - An object.

Returns:

True if Object is a Regressor with the same dimension names, false otherwise.

hashCode

public int hashCode()

Regressor's hashcode is based on the hash of the dimension names.

It's cached on first access.

Overrides:

hashCode in class Object

Returns:

A hashcode.

getDimensionNamesString

public String getDimensionNamesString()

Returns a comma separated list of the dimension names.

Returns:

The dimension names comma separated.

getDimensionNamesString

public String getDimensionNamesString(char separator)

Returns a delimiter separated list of the dimension names.

Parameters:

separator - The separator to use.

Returns:

The dimension names.

extractNames

public static String[] extractNames(OutputInfo<Regressor> info)

Extracts the names from the supplied Regressor domain in their canonical order.

Parameters:

info - The OutputInfo to use.

Returns:

The dimension names from this domain.

parseString

public static Regressor parseString(String s)

Parses a string of the form:

 dimension-name=output,...,dimension-name=output
 

or

 output,...,output
 

where output must be readable by Double.parseDouble(java.lang.String). If there are no dimension names specified then they are automatically generated based on the position in the string.

Parameters:

s - The string form of a multiple regressor.

Returns:

A regressor parsed from the input string.

parseString

public static Regressor parseString(String s,
                                    char splitChar)

Parses a string of the form:

 dimension-name=output<splitChar>...<splitChar>dimension-name=output
 

or

 output<splitChar>...<splitChar>output
 

where output must be readable by Double.parseDouble(java.lang.String). If there are no dimension names specified then they are automatically generated based on the position in the string.

Parameters:

s - The string form of a regressor.

splitChar - The char to split on.

Returns:

A regressor parsed from the input string.

parseElement

public static com.oracle.labs.mlrg.olcut.util.Pair<String,Double> parseElement(int idx,
                                                                               String s)

Parses a string of the form:

 dimension-name=output-double
 

or

 output-double
 

where the output must be readable by Double.parseDouble(java.lang.String). If there is no dimension name then one is constructed by appending idx to DEFAULT_NAME.

Parameters:

idx - The index of this string in a list.

s - The string form of a single dimension from a regressor.

Returns:

A tuple representing the dimension name and the value.

createFromPairList

public static Regressor createFromPairList(List<com.oracle.labs.mlrg.olcut.util.Pair<String,Double>> dimensions)

Creates a Regressor from a list of dimension tuples.

Parameters:

dimensions - The dimensions to use.

Returns:

A Regressor representing these dimensions.