OptimizationTargetInterface Struct Reference

The interface to be provided by functions to be optimized. More...

Public Member Functions
virtual Tensor< double >	gradient (const Tensor< double > &x)
	Should return the derivative of the function.
virtual bool	provides_gradient () const =0
	Should return true if the gradient is implemented.
double	test_gradient (Tensor< double > &x, double value_precision, bool doprint=true)
	Numerical test of the derivative ... optionally prints to stdout, returns max abs error.
virtual double	value (const Tensor< double > &x)=0
	Should return the value of the objective function.
virtual void	value_and_gradient (const Tensor< double > &x, double &value, Tensor< double > &gradient)
	Reimplement if more efficient to evaluate both value and gradient in one call.

Detailed Description

The interface to be provided by functions to be optimized.

Member Function Documentation

gradient()

virtual Tensor< double > OptimizationTargetInterface::gradient ( const Tensor< double > &  x )

inlinevirtual

Should return the derivative of the function.

Referenced by test_gradient(), and value_and_gradient().

provides_gradient()

virtual bool OptimizationTargetInterface::provides_gradient ( ) const

pure virtual

Should return true if the gradient is implemented.

test_gradient()

double OptimizationTargetInterface::test_gradient ( Tensor< double > &  x, double  value_precision, bool  doprint = true  )

inline

Numerical test of the derivative ... optionally prints to stdout, returns max abs error.

References madness::abs(), madness::BaseTensor::dim(), gradient(), max, pow(), and value().

value()

virtual double OptimizationTargetInterface::value ( const Tensor< double > &  x )

pure virtual

Should return the value of the objective function.

Referenced by test_gradient(), and value_and_gradient().

value_and_gradient()

virtual void OptimizationTargetInterface::value_and_gradient ( const Tensor< double > &  x, double &  value, Tensor< double > &  gradient  )

inlinevirtual

Reimplement if more efficient to evaluate both value and gradient in one call.

References gradient(), and value().

---

The documentation for this struct was generated from the following file:

• kain.cc