BFGSOptimiser (PULsE 1.98 API)

java.lang.Object
- pulse.util.UpwardsNavigable
- - pulse.util.Group
  - - pulse.util.Accessible
    - - pulse.util.PropertyHolder
      - pulse.search.direction.PathOptimiser
        
        pulse.search.direction.GradientBasedOptimiser
        
        pulse.search.direction.CompositePathOptimiser
        
        pulse.search.direction.BFGSOptimiser

All Implemented Interfaces:

Serializable, Descriptive, Reflexive
```
public class BFGSOptimiser
extends CompositePathOptimiser
```
The 'advanced' PathSolver implementing the variable-metric (quasi-Newton) search method.
The latter does not only rely on the gradient (first derivatives) of the target function, as commonly used in simpler optimisation methods, such as the steepest descent method, but also accounts for the second-order derivatives. This leads to an additional term in the equation defining the minimum direction. This term is called the 'Hessian' matrix, which is calculated approximately using the BFGS formula. Note that the initial value for the 'Hessian' matrix is an identity matrix. It is recommended to use this PathSolver in combination with the WolfeSolver.

See Also:

Wikipedia page, WolfeOptimiser, Serialized Form

Field Summary
- Fields inherited from class pulse.search.direction.CompositePathOptimiser
  EPS, MAX_FAILED_ATTEMPTS

Method Summary

All Methods Static Methods Instance Methods Concrete Methods
Modifier and Type	Method	Description
`static BFGSOptimiser`	`getInstance()`	This class uses a singleton pattern, meaning there is only instance of this class.
`void`	`prepare(GeneralTask task)`	Calculated the gradient at the end of this step.
`String`	`toString()`

Methods inherited from class pulse.search.direction.CompositePathOptimiser
getLinearOptimiserDescriptor, getLinearSolver, initState, iteration, listedTypes, setLinearSolver

Methods inherited from class pulse.search.direction.GradientBasedOptimiser
dx, getGradientResolution, getGradientStep, gradient, listedKeywords, reset, set, setGradientResolution

Methods inherited from class pulse.search.direction.PathOptimiser
compatibleWith, genericProperties, getErrorTolerance, getMaxIterations, getSolver, ignoreSiblings, setErrorTolerance, setInstance, setMaxIterations, setSolver, update

Methods inherited from class pulse.util.PropertyHolder
addListener, areDetailsHidden, data, describe, firePropertyChanged, getDescriptor, getListeners, getPrefix, initListeners, isListedNumericType, isListedParameter, numericData, parameterListChanged, removeListeners, setPrefix, updateProperties, updateProperty

Methods inherited from class pulse.util.Accessible
accessibleChildren, genericProperty, numericProperties, numericProperty, property, update

Methods inherited from class pulse.util.Group
access, children, contents, getSimpleName, subgroups

Methods inherited from class pulse.util.UpwardsNavigable
addHierarchyListener, getHierarchyListeners, getParent, identify, removeHierarchyListener, removeHierarchyListeners, setParent, specificAncestor, tellParent

Methods inherited from class java.lang.Object
clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

- Method Detail
  - prepare
```
public void prepare(GeneralTask task)
             throws SolverException
```
    Calculated the gradient at the end of this step. Invokes hessian(...) to calculate the Hessian matrix at the <i>k</i>+1 step using the g_k and g_k+1 gradient values, the previously calculated Hessian matrix on step k, and the result of the linear search α_k+1.
    
    Specified by:
    
    prepare in class PathOptimiser
    
    Parameters:
    
    task - the SearchTask undergoing optimisation
    
    Throws:
    
    SolverException
  - toString
```
public String toString()
```
    Overrides:
    
    toString in class PathOptimiser
  - getInstance
```
public static BFGSOptimiser getInstance()
```
    This class uses a singleton pattern, meaning there is only instance of this class.
    
    Returns:
    
    the single (static) instance of this class