de.dfki.mycbr.core.similarity
Class DoubleFct

java.lang.Object
  java.util.Observable
      de.dfki.mycbr.core.similarity.NumberFct
          de.dfki.mycbr.core.similarity.DoubleFct

All Implemented Interfaces:

ISimFct, java.util.Observer

public class DoubleFct
extends NumberFct

Calculates the similarity of DoubleAttribute objects. There are four modes which can be chosen for the similarity function:

• CONSTANT: the similarity is always a constant value
• STEP AT: for a distance smaller than the specified value the similarity is 0.00. For the other values it is 1.00
• POLINOMIAL WITH: stretches/compresses the line described by y = x/(max-min) + 1 ( for case < query ) resp. y = - (x/(max-min)) +1 ( for case > query ) by taking y to the power of the given parameter
• SMOOTH STEP AT: similar to STEP AT but with a smooth step, the function is computed as the logistic function 1/(1 + exp((x-param)*100/(max-min))) ( cp. sigmoid-curve or s-curve)

You have to configure the similarity for two cases:

• query < case: referred to as the right part of the function
• case < query: referred to as the left part of the function

If the function is symmetric, the configurations should be the same. The similarity of a value to itself is assumed to be 1.00 in any case.

Author:

myCBR Team

Constructor Summary

DoubleFct(Project prj, DoubleDesc desc, java.lang.String name)
Constructor should only be called by FunctionContainer.

Method Summary

Similarity	calculateSimilarity(Attribute a1, Attribute a2) Returns the similarity of the given attributes.
Similarity	calculateSimilarity(java.lang.Double value1, java.lang.Double value2) Calculates the similarity of the given Floats.
void	clone(AttributeDesc descNEW, boolean active) Creates a new function which is the same function as this but with a different description, namely descNEW.
boolean	fitsDistance(double x)
double	getDiff()
double	getFunctionParameterL() Gets the function's parameter for case < query to value.
double	getFunctionParameterR() Gets the function's parameter for query < case to value.
NumberConfig	getFunctionTypeL()
NumberConfig	getFunctionTypeR()
double	getMax()
double	getMaxForQuotient()
double	getMin()
void	setDistanceFct(DistanceConfig df) Sets the mode of this function to mode
boolean	setFunctionParameterL(double value) Sets the function's parameter for query < case to value.
boolean	setFunctionParameterR(double value) Sets the function's parameter for query < case to value.
void	setFunctionTypeL(NumberConfig type) Sets the type of the function for case < query.
void	setFunctionTypeR(NumberConfig type) Sets the type of the function for query < case.
void	setMaxForQuotient(double max)
void	update(java.util.Observable arg0, java.lang.Object arg1)
void	updateFunctionParams()

Methods inherited from class de.dfki.mycbr.core.similarity.NumberFct

getDesc, getDistanceFct, getMultipleConfig, getName, getProject, isSymmetric, setMultipleConfig, setName, setSymmetric

Methods inherited from class java.util.Observable

addObserver, countObservers, deleteObserver, deleteObservers, hasChanged, notifyObservers, notifyObservers

Methods inherited from class java.lang.Object

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

Constructor Detail

DoubleFct

public DoubleFct(Project prj,
                 DoubleDesc desc,
                 java.lang.String name)

Constructor should only be called by FunctionContainer. Initializes this with the given description.

Parameters:

desc - the description of the attributes this function can be applied to

Method Detail

calculateSimilarity

public Similarity calculateSimilarity(Attribute a1,
                                      Attribute a2)
                               throws java.lang.Exception

Returns the similarity of the given attributes. Returns invalid similarity if an error occurs. The similarity computation depends on the type of the function and it's parameter. We refer to the right part of the function if case < query and to the left part otherwise. The possible computations are:

• CONSTANT: the similarity is always a constant value
• STEP AT: for a distance smaller than the specified value the similarity is 0.00. For the other values it is 1.00
• POLINOMIAL WITH: stretches/compresses the line described by y = x/(max-min) + 1 ( for case < query ) resp. y = - (x/(max-min)) +1 ( for case > query ) by taking y to the power of the given parameter
• SMOOTH STEP AT: similar to STEP AT but with a smooth step, the function is computed as the logistic function 1/(1 + exp((x-param)*100/(max-min))) ( cp. sigmoid-curve or s-curve)

Parameters:

a1 - the first value

a2 - the second value

Returns:

Similarity of the given attributes, invalid similarity if an error occurs.

Throws:

java.lang.Exception

calculateSimilarity

public Similarity calculateSimilarity(java.lang.Double value1,
                                      java.lang.Double value2)
                               throws java.lang.Exception

Calculates the similarity of the given Floats. Calls calculateSimilarity(Attribute, Attribute). Be aware that calculateSimilarity(value1, values2) might differ from calculateSimilarity(value2, values1). When calculating similarities, the first attribute is always referred to as the query value, the second parameter as the case value.

Parameters:

value1 - the first attribute to be compared

value2 - the second attribute to be compared

Returns:

similarity of the given attributes

Throws:

java.lang.Exception

setFunctionParameterR

public boolean setFunctionParameterR(double value)

Sets the function's parameter for query < case to value. The function parameter depends on the type of the function. The parameter specifies:

• CONSTANT: the constant similarity value
• STEP AT: the distance value where the step should be (every distance greater or equal than this value has similarity 1.00, else 0.00)
• POLINOMIAL WITH: the constant with which the function gets compressed or stretched
• SMOOTH STEP AT: the distance value where the smooth step should be

If the parameter for "smooth step" or "step at" does not fit the range given by max/min, the parameter will not be updated. If the parameter for "constant" is not within the range (0.0,1.0), the parameter will not be updated. Or if the parameter for "polynomial with" is smaller than 0 the parameter will not be updated. Be aware to set the function type before setting the parameter, because the parameter depends on the current function type.

Parameters:

value - the function parameter for query < case

Returns:

true, if the parameter was successfully updated, false otherwise

setFunctionParameterL

public boolean setFunctionParameterL(double value)

Sets the function's parameter for query < case to value. The function parameter depends on the type of the function. The parameter specifies:

• CONSTANT: the constant similarity value
• STEP AT: the distance value where the step should be (every distance greater or equal than this value has similarity 1.00, else 0.00)
• POLINOMIAL WITH: the constant with which the function gets compressed or stretched
• SMOOTH STEP AT: the distance value where the smooth step should be

If the parameter for "smooth step" or "step at" does not fit the range given by max/min, the parameter will not be updated. If the parameter for "constant" is not within the range (0.0,1.0), the parameter will not be updated. Or if the parameter for "polynomial with" is smaller than 0 the parameter will not be updated. Be aware to set the function type before setting the parameter, because the parameter depends on the current function type.

Parameters:

value - the function parameter for query < case

Returns:

true, if the parameter was successfully updated, false otherwise

getFunctionParameterR

public double getFunctionParameterR()

Gets the function's parameter for query < case to value. The function parameter depends on the type of the function. The parameter specifies:

• CONSTANT: the constant similarity value
• STEP AT: the distance value where the step should be (every distance greater or equal than this value has similarity 1.00, else 0.00)
• POLINOMIAL WITH: the constant with which the function gets compressed or stretched
• SMOOTH STEP AT: the distance value where the smooth step should be

Returns:

the function parameter for query < case

getFunctionParameterL

public double getFunctionParameterL()

Gets the function's parameter for case < query to value. The function parameter depends on the type of the function. The parameter specifies:

• CONSTANT: the constant similarity value
• STEP AT: the distance value where the step should be (every distance greater or equal than this value has similarity 1.00, else 0.00)
• POLINOMIAL WITH: the constant with which the function gets compressed or stretched
• SMOOTH STEP AT: the distance value where the smooth step should be

Returns:

the function parameter for case < query

setFunctionTypeR

public void setFunctionTypeR(NumberConfig type)

Sets the type of the function for query < case. Updates function parameter if necessary. Be aware to set the function type before setting the parameter, because the parameter depends on the current function type.

Parameters:

type - the type of the function

setFunctionTypeL

public void setFunctionTypeL(NumberConfig type)

Sets the type of the function for case < query. Updates function parameter if necessary. Be aware to set the function type before setting the parameter, because the parameter depends on the current function type.

Parameters:

type - the type of the function

getFunctionTypeL

public NumberConfig getFunctionTypeL()

getFunctionTypeR

public NumberConfig getFunctionTypeR()

getMin

public double getMin()

getMax

public double getMax()

getDiff

public double getDiff()

setDistanceFct

public void setDistanceFct(DistanceConfig df)

Description copied from class: NumberFct

Sets the mode of this function to mode

Overrides:

setDistanceFct in class NumberFct

Parameters:

df - the new mode of this function

clone

public void clone(AttributeDesc descNEW,
                  boolean active)

Description copied from interface: ISimFct

Creates a new function which is the same function as this but with a different description, namely descNEW. Assumption: this function fits the description of descNEW.

update

public void update(java.util.Observable arg0,
                   java.lang.Object arg1)

updateFunctionParams

public void updateFunctionParams()

fitsDistance

public boolean fitsDistance(double x)

Parameters:

x - a value to be checked

Returns:

true, if x is still in the domain defined by desc wrt to the given distance function

getMaxForQuotient

public double getMaxForQuotient()

setMaxForQuotient

public void setMaxForQuotient(double max)