cs2.introml

Class HiddenPatternDataSet

java.lang.Object

cs2.introml.DataSet

cs2.introml.HiddenPatternDataSet

All Implemented Interfaces:

java.lang.Iterable<Example>

public class HiddenPatternDataSet
extends DataSet

A dataset based upon a set of patterns, each a result object and a set of feature number/value pairs. (This simulates an easy pattern hiding in the data.) If provided a random seed, all remaining features are selected from a uniform random distribution (with configurable bounds). Otherwise, all other features are zero. The dataset deterministically cycles through the patterns, producing one example per pattern. There cannot be more patterns than examples. If there are more examples than patterns, the dataset cycles back through the patterns as many times as necessary. If using a random distribution, each non-patterned feature will be randomly selected between subsequent cycles of the pattern (i.e. two examples of the same pattern will most likely have different values for each non-patterned feature).

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	HiddenPatternDataSet.PatternException Exception class for errors adding patterns

Constructor Summary

Constructors

Constructor and Description
HiddenPatternDataSet(int numFeatures, int numExamples) Initializes the data set using 0's whenever a pattern does not specify a feature number
HiddenPatternDataSet(int numFeatures, int numExamples, long rngSeed) Initializes the data set using random values [0,1] whenever a pattern does not specify a feature number
HiddenPatternDataSet(int numFeatures, int numExamples, long rngSeed, double minRand, double maxRand) Initializes the data set using random values [min,max] whenever a pattern does not specify a feature number

Method Summary

Modifier and Type	Method and Description
HiddenPatternDataSet	addPattern(java.lang.Object result, java.util.Map<java.lang.Integer,java.lang.Double> pattern) Adds a pattern to the dataset
int	getNumExamples() Gets the number of examples in the dataset (might change over time).
java.util.Iterator<Example>	iterator()

Methods inherited from class cs2.introml.DataSet

getNumFeatures, toString

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, wait, wait, wait

Methods inherited from interface java.lang.Iterable

forEach, spliterator

Constructor Detail

HiddenPatternDataSet

public HiddenPatternDataSet(int numFeatures,
                            int numExamples)

Initializes the data set using 0's whenever a pattern does not specify a feature number

Parameters:

numFeatures - number of features

numExamples - number of examples

HiddenPatternDataSet

public HiddenPatternDataSet(int numFeatures,
                            int numExamples,
                            long rngSeed)

Initializes the data set using random values [0,1] whenever a pattern does not specify a feature number

Parameters:

numFeatures - number of features

numExamples - number of examples

rngSeed - seed for the random number generator

HiddenPatternDataSet

public HiddenPatternDataSet(int numFeatures,
                            int numExamples,
                            long rngSeed,
                            double minRand,
                            double maxRand)

Initializes the data set using random values [min,max] whenever a pattern does not specify a feature number

Parameters:

numFeatures - number of features

numExamples - number of examples

rngSeed - seed for the random number generator

minRand - minimum value for the random feature values

maxRand - maximum value for the random feature values

Method Detail

getNumExamples

public int getNumExamples()

Description copied from class: DataSet

Gets the number of examples in the dataset (might change over time).

Specified by:

getNumExamples in class DataSet

Returns:

number of examples

addPattern

public HiddenPatternDataSet addPattern(java.lang.Object result,
                                       java.util.Map<java.lang.Integer,java.lang.Double> pattern)
                                throws HiddenPatternDataSet.PatternException

Adds a pattern to the dataset

Parameters:

result - pattern result object (only one pattern per result)

pattern - set of feature number/value pairs

Returns:

this dataset (may be useful for chaining)

Throws:

HiddenPatternDataSet.PatternException - errors in the pattern (e.g. existing pattern for result, more patterns than examples, invalid feature number, null feature value)

iterator

public java.util.Iterator<Example> iterator()