RDPickers::HierarchicalClusterPicker Class Reference

Diversity picker based on hierarchical clustering. More...

#include <HierarchicalClusterPicker.h>

Inheritance diagram for RDPickers::HierarchicalClusterPicker:

List of all members.

Public Types
enum	ClusterMethod {   WARD = 1, SLINK = 2, CLINK = 3, UPGMA = 4,   MCQUITTY = 5, GOWER = 6, CENTROID = 7 }
	The type of hierarchical clustering algorithm to use. More...
Public Member Functions
	HierarchicalClusterPicker (ClusterMethod clusterMethod)
	Constructor - takes a ClusterMethod as an argument.
RDKit::INT_VECT	pick (const double *distMat, unsigned int poolSize, unsigned int pickSize) const
	This is the function that does the picking.
RDKit::VECT_INT_VECT	cluster (const double *distMat, unsigned int poolSize, unsigned int pickSize) const
	This is the function that does the clustering of the items - used by the picker.

---

Detailed Description

Diversity picker based on hierarchical clustering.

This class inherits from DistPicker since it uses the distance matrix for diversity picking. The clustering itself is done using the Murtagh code in $RDBASE/Code/ML/Cluster/Mutagh/

Definition at line 20 of file HierarchicalClusterPicker.h.

---

Member Enumeration Documentation

enum RDPickers::HierarchicalClusterPicker::ClusterMethod

The type of hierarchical clustering algorithm to use.

Enumerator:

WARD
SLINK
CLINK
UPGMA
MCQUITTY
GOWER
CENTROID

Definition at line 25 of file HierarchicalClusterPicker.h.

---

Constructor & Destructor Documentation

RDPickers::HierarchicalClusterPicker::HierarchicalClusterPicker

(

ClusterMethod

clusterMethod

)

[inline, explicit]

Constructor - takes a ClusterMethod as an argument.

Sets the hierarch clustering method

Definition at line 38 of file HierarchicalClusterPicker.h.

---

Member Function Documentation

RDKit::INT_VECT RDPickers::HierarchicalClusterPicker::pick	(	const double *	distMat,
		unsigned int	poolSize,
		unsigned int	pickSize
	)			const [virtual]

This is the function that does the picking.

Here is how the algorithm works
FIX: Supply reference

• The entire pool is clustered using the distance matrix using one of the hierachical clustering method (specified via the constructor).
  
• Starting with the individaul items in the pool, clusters are merged based on the output from clustering method.
  
• The merging is stopped when the number of clusters is same as the number of picks.
• For each item in a cluster the sum of square of the distances to the rest of of the items (in the cluster) is computed. The item with the smallest of values is picked as a representative of the cluster. Basically trying to pick the item closest to the centroid of the cluster.

Parameters:

	distMat	- distance matrix - a vector of double. It is assumed that only the lower triangle element of the matrix are supplied in a 1D array NOTE: this matrix WILL BE ALTERED during the picking
	poolSize	- the size of the pool to pick the items from. It is assumed that the distance matrix above contains the right number of elements; i.e. poolSize*(poolSize-1)
	pickSize	- the number items to pick from pool (<= poolSize)

Implements RDPickers::DistPicker.

RDKit::VECT_INT_VECT RDPickers::HierarchicalClusterPicker::cluster	(	const double *	distMat,
		unsigned int	poolSize,
		unsigned int	pickSize
	)			const

This is the function that does the clustering of the items - used by the picker.

ARGUMENTS:

Parameters:

	distMat	- distance matrix - a vector of double. It is assumed that only the lower triangle element of the matrix are supplied in a 1D array NOTE: this matrix WILL BE ALTERED during the picking
	poolSize	- the size of the pool to pick the items from. It is assumed that the distance matrix above contains the right number of elements; i.e. poolSize*(poolSize-1)
	pickSize	- the number clusters to divide the pool into (<= poolSize)

---

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

• HierarchicalClusterPicker.h

---