umontreal.iro.lecuyer.util

Class RatioFunction

java.lang.Object

umontreal.iro.lecuyer.util.RatioFunction

All Implemented Interfaces:

MultivariateFunction

public class RatioFunction
extends java.lang.Object
implements MultivariateFunction

Represents a function computing a ratio of two values.

Constructor Summary

Constructors

Constructor and Description
RatioFunction() Constructs a new ratio function.
RatioFunction(double zeroOverZero) Constructs a new ratio function that returns zeroOverZero for the special case of 0/0.

Method Summary

Modifier and Type	Method and Description
double	evaluate(double... x) Computes the function g(x) for the vector x.
double	evaluateGradient(int i, double... x) Computes ∂g(x)/∂xi, the derivative of g(x) with respect to xi.
int	getDimension() Returns d, the dimension of the function computed by this implementation.
double	getZeroOverZeroValue() Returns the value returned by evaluate in the case where the 0/0 function is calculated.
void	setZeroOverZeroValue(double zeroOverZero) Sets the value returned by evaluate for the undefined function 0/0 to zeroOverZero.

Methods inherited from class java.lang.Object

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

Constructor Detail

RatioFunction

public RatioFunction()

Constructs a new ratio function.

RatioFunction

public RatioFunction(double zeroOverZero)

Constructs a new ratio function that returns zeroOverZero for the special case of 0/0. See the getZeroOverZeroValue method for more information. The default value of zeroOverZero is Double.NaN.

Parameters:

zeroOverZero - the value for 0/0.

Method Detail

getZeroOverZeroValue

public double getZeroOverZeroValue()

Returns the value returned by evaluate in the case where the 0/0 function is calculated. The default value for 0/0 is Double.NaN.

Generally, 0/0 is undefined, and therefore associated with the Double.NaN constant, meaning not-a-number. However, in certain applications, it can be defined differently to accomodate some special cases. For exemple, in a queueing system, if there are no arrivals, no customers are served, lost, queued, etc. As a result, many performance measures of interest turn out to be 0/0. Specifically, the loss probability, i.e., the ratio of lost customers over the number of arrivals, should be 0 if there is no arrival; in this case, 0/0 means 0. On the other hand, the service level, i.e., the fraction of customers waiting less than a fixed threshold, could be fixed to 1 if there is no arrival.

Returns:

the value for 0/0.

setZeroOverZeroValue

public void setZeroOverZeroValue(double zeroOverZero)

Sets the value returned by evaluate for the undefined function 0/0 to zeroOverZero. See getZeroOverZeroValue for more information.

Parameters:

zeroOverZero - the new value for 0/0.

getDimension

public int getDimension()

Description copied from interface: MultivariateFunction

Returns d, the dimension of the function computed by this implementation. If the dimension is not fixed, this method must return a negative value.

Specified by:

getDimension in interface MultivariateFunction

Returns:

the dimension.

evaluate

public double evaluate(double... x)

Description copied from interface: MultivariateFunction

Computes the function g(x) for the vector x. The length of the given array must correspond to the dimension of this function. The method must compute and return the result of the function without modifying the elements in x since the array can be reused for further computation.

Specified by:

evaluate in interface MultivariateFunction

Parameters:

x - a vector x.

Returns:

the value of g(x).

evaluateGradient

public double evaluateGradient(int i,
                               double... x)

Description copied from interface: MultivariateFunction

Computes ∂g(x)/∂x_i, the derivative of g(x) with respect to x_i. The length of the given array must correspond to the dimension of this function. The method must compute and return the result of the derivative without modifying the elements in x since the array can be reused for further computations, e.g., the gradient ∇g(x).

Specified by:

evaluateGradient in interface MultivariateFunction

Parameters:

i - the variable to derive with respect to.

x - a vector x.

Returns:

the value of the partial derivative.