Reinforcement Learning Toolbox 2.0 last updated:

General Documentation Manual Tutorial Class Reference Master Thesis Examples Related Papers Downloads Links News mailto:webmaster Main Page     Class Hierarchy   Compound List   File List   Compound Members   File Members CRBFFeatureCalculator Class Reference Represents a normalized RBF network. More... #include <clinearfafeaturecalculator.h> Inheritance diagram for CRBFFeatureCalculator: List of all members. Public Member Functions   CRBFFeatureCalculator (unsigned int numDim, unsigned int dimensions[], unsigned int partitions[], double offsets[], double sigma[])   CRBFFeatureCalculator (unsigned int numDim, unsigned int dimensions[], unsigned int partitions[], double offsets[], double sigma[], unsigned int areaSize[])   ~CRBFFeatureCalculator () Protected Member Functions virtual double  getFeatureFactor (CState *state, ColumnVector *featPos)   Calculates the feature factor with the RBF function. virtual void  initAreaSize ()   set the areaSize array Protected Attributes double *  sigma double  sigmaMaxSize Detailed Description Represents a normalized RBF network. This class lays a grid of RBF-Functions over a specified sub-space of the model space. You can specify which dimensions you want to use (dimensions array), how much partitions you need for each dimension, an offset for each dimension which is added to the RBF centers and the sigmas for each dimension. When talking about grids, the RBF center is always in the center of the partition. The size of the feature space is the product of all numbers of partitions. This class always works with the normalized continuous state variable values (scaled to the intervall [0, 1], so consider that when choosing your offsets and sigmas. The number of active states for each dimension is choosen according the sigma value. All features are active that are within a range of 2 * sigma. So if you want only the neighbors to be active too, use the thumbrule sigma[i] = 1 / (2 * numPartitions[i] ), so you get 3 active features per dimension. Constructor & Destructor Documentation CRBFFeatureCalculator::CRBFFeatureCalculator (  unsigned int  numDim, unsigned int  dimensions[], unsigned int  partitions[], double  offsets[], double  sigma[] )  CRBFFeatureCalculator::CRBFFeatureCalculator (  unsigned int  numDim, unsigned int  dimensions[], unsigned int  partitions[], double  offsets[], double  sigma[], unsigned int  areaSize[] )  CRBFFeatureCalculator::~CRBFFeatureCalculator (   )  Member Function Documentation virtual double CRBFFeatureCalculator::getFeatureFactor (  CState *  state, ColumnVector *  featPos )  [protected, virtual]   Calculates the feature factor with the RBF function. Implements CLinearMultiFeatureCalculator. virtual void CRBFFeatureCalculator::initAreaSize (   )  [protected, virtual]   set the areaSize array The number of active states for each dimension is choosen according the sigma value. All features are active that are within a range of 2 * sigma. So if you want only the neighbors to be active too, use the thumbrule sigma[i] = 1 / (2 * numPartitions[i] ), so you get 3 active features per dimension. Reimplemented from CLinearMultiFeatureCalculator. Member Data Documentation double* CRBFFeatureCalculator::sigma [protected] double CRBFFeatureCalculator::sigmaMaxSize [protected] The documentation for this class was generated from the following file: clinearfafeaturecalculator.h