org.apache.commons.math3.optim.univariate
public class BrentOptimizer extends UnivariateOptimizer
(lo, hi)
, this class
finds an approximation x
to the point at which the function
attains its minimum.
It implements Richard Brent's algorithm (from his book "Algorithms for
Minimization without Derivatives", p. 79) for finding minima of real
univariate functions.
Modifier and Type | Field and Description |
---|---|
private double |
absoluteThreshold
Absolute threshold.
|
private static double |
GOLDEN_SECTION
Golden section.
|
private static double |
MIN_RELATIVE_TOLERANCE
Minimum relative tolerance.
|
private double |
relativeThreshold
Relative threshold.
|
evaluations, iterations
Constructor and Description |
---|
BrentOptimizer(double rel,
double abs)
The arguments are used for implementing the original stopping criterion
of Brent's algorithm.
|
BrentOptimizer(double rel,
double abs,
ConvergenceChecker<UnivariatePointValuePair> checker)
The arguments are used implement the original stopping criterion
of Brent's algorithm.
|
Modifier and Type | Method and Description |
---|---|
private UnivariatePointValuePair |
best(UnivariatePointValuePair a,
UnivariatePointValuePair b,
boolean isMinim)
Selects the best of two points.
|
protected UnivariatePointValuePair |
doOptimize()
Performs the bulk of the optimization algorithm.
|
computeObjectiveValue, getGoalType, getMax, getMin, getStartValue, optimize
getConvergenceChecker, getEvaluations, getIterations, getMaxEvaluations, getMaxIterations, incrementEvaluationCount, incrementIterationCount
private static final double GOLDEN_SECTION
private static final double MIN_RELATIVE_TOLERANCE
private final double relativeThreshold
private final double absoluteThreshold
public BrentOptimizer(double rel, double abs, ConvergenceChecker<UnivariatePointValuePair> checker)
abs
and rel
define a tolerance
tol = rel |x| + abs
. rel
should be no smaller than
2 macheps and preferably not much less than sqrt(macheps),
where macheps is the relative machine precision. abs
must
be positive.rel
- Relative threshold.abs
- Absolute threshold.checker
- Additional, user-defined, convergence checking
procedure.NotStrictlyPositiveException
- if abs <= 0
.NumberIsTooSmallException
- if rel < 2 * Math.ulp(1d)
.public BrentOptimizer(double rel, double abs)
abs
and rel
define a tolerance
tol = rel |x| + abs
. rel
should be no smaller than
2 macheps and preferably not much less than sqrt(macheps),
where macheps is the relative machine precision. abs
must
be positive.rel
- Relative threshold.abs
- Absolute threshold.NotStrictlyPositiveException
- if abs <= 0
.NumberIsTooSmallException
- if rel < 2 * Math.ulp(1d)
.protected UnivariatePointValuePair doOptimize()
doOptimize
in class BaseOptimizer<UnivariatePointValuePair>
private UnivariatePointValuePair best(UnivariatePointValuePair a, UnivariatePointValuePair b, boolean isMinim)
a
- Point and value.b
- Point and value.isMinim
- true
if the selected point must be the one with
the lowest value.null
if a
and b
are
both null
. When a
and b
have the same function
value, a
is returned.Copyright (c) 2003-2013 Apache Software Foundation