Defines the interface for a feature mapping $\phi : \mathbb{R}^n \rightarrow \mathbb{R}^m$ . More...

Inheritance diagram for drwnFeatureMap:

Public Member Functions
	drwnFeatureMap ()
	default constructor

	drwnFeatureMap (int nFeatures)
	construct with known number of input features

virtual	~drwnFeatureMap ()
	destructor

virtual drwnFeatureMap *	clone () const =0
	returns a copy of the class usually implemented as virtual Foo* clone() { return new Foo(*this); }

int	numFeatures () const
	returns the number of features in the input space

virtual int	numParameters () const =0
	returns the number of features in the output space

virtual void	initialize (int nFeatures)
	initialize number of (intput) features

virtual vector< double >	operator() (const vector< double > &x) const =0
	returns $\phi(x) \in \mathbb{R}^m$ where $x \in \mathbb{R}^n$

virtual double	dot (const vector< double > &theta, const vector< double > &x) const
	returns the dot product $\theta^T \phi(x)$ (default behaviour is very inefficient)

virtual void	mac (vector< double > &theta, const vector< double > &x, double alpha) const
	provides multiply-accumulate operation $\theta = \theta + \alpha \phi(x)$ (default behaviour is very inefficient)

Public Member Functions inherited from drwnTypeable
virtual const char *	type () const =0
	returns object type as a string (e.g., Foo::type() { return "Foo"; })

Detailed Description

Defines the interface for a feature mapping $\phi : \mathbb{R}^n \rightarrow \mathbb{R}^m$ .

Warning: The default behaviour of all virtual functions are very inefficient and should be replaced in the derived classes. They are provided for testing during development, requiring the programmer to only implement the operator()(const vector<double>& x) function.

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

Public Member Functions