Habbersett Contrib Library

This page is a listing of the entire contents of this library for IDL. This listing is the long version. Viewing the much more compact listing may be handier.

[Go Back to Main IDL Libraries Search Page]


Last modified: Thu Dec 21 21:32:50 2000.

List of Routines


Routine Descriptions

NN_CLUST

[Next Routine] [List of Routines]
 NAME:		nn_clust.pro

 PURPOSE:	tag events with cluster #

 CATEGORY:	data processing - cluster analysis

 CALLING SEQUENCE:	result = NN_CLUST(data,weights,[N_PARAM=n_param],
			[N_EVENT=n_event],[N_CLUST=n_clust])

 INPUTS: data = data set to be clustered, weights = array of weights which
	  describe cluster centers (produced by the function NN_LEARN)

 OPTIONAL INPUT PARAMETERS:
	n_param = # of parameters in data set - 1st dimension of data set
	n_event = # of events in data set - 2nd dimension of the data set
	n_clust = # of clusters to find - arbitrary

 KEYWORD PARAMETERS:	none

 OUTPUTS:	result = integer array of cluster assignment of each event

 OPTIONAL OUTPUT PARAMETERS:	none

 COMMON BLOCKS:	none

 SIDE EFFECTS:	none (?)

 RESTRICTIONS:	Read the header of 'nn_learn.pro'.

 PROCEDURE: The data set is processed against the weights array and a cluster
	     # assignment is made for each event in the input data set based
	     on the weights array which describe the center of each cluster.

 MODIFICATION HISTORY:	initial algorithm: Mark Naiver (Univ of Texas - Austin)
	Date last modified ==>  1 March 93 : RCH [LANL]
	Contact: Robb Habbersett (505/667-0296 or robb@big-geek.lanl.gov)

(See /host/bluemoon/usr2/idllib/user_contrib/habbersett/nn_clust.pro)


NN_LEARN

[Previous Routine] [List of Routines]
 NAME:		nn_learn.pro

 PURPOSE:	Learning step to cluster data using neural network techniques.

 CATEGORY:	Data processing - cluster analysis

 CALLING SEQUENCE:	weights = nn_learn(data,max_val,[BLR=blr],[ELR=elr],
	[N_EVENT=n_event],[N_PASS=n_pass],[N_PARAM=n_param],[N_CLUST=n_clust]

 INPUTS:	data = data set to be clustered, max_val = maximum range of each
		parameter in the data set (to normalize the weights).

 OPTIONAL INPUT PARAMETERS:
	blr = begining learning rate, elr = ending learning rate
	n_pass = # of iterations of the learning pass
	n_param = # of parameters in data set - 1st dimension of data set
	n_event = # of events in learning set - 2nd dimension of data set
	n_clust = # of clusters to find - arbitrary (?)

 KEYWORD PARAMETERS:	none

 OUTPUTS:	result = an array of weights describing the cluster centers.

 OPTIONAL OUTPUT PARAMETERS:	none

 COMMON BLOCKS:	none

 SIDE EFFECTS: This approach has an inherent weakness in that it must be
		set to find a specific number of clusters; It will find
		that number of clusters in the data set - regardless.

 RESTRICTIONS:	This routine has not been rigorously tested on different types
		of data. It "appears" to work on flow cytometry data.

 PROCEDURE:	A limited subset of a larger data set is presented to this
		routine as a training set to condition the neural network. The
		result is a set of weights which describe the centers of the
		resolved clusters.

 MODIFICATION HISTORY:	Initial algorithm: Mark Naiver (Univ of Texas - Austin)
	Date last modified ==>  1 March 93 : RCH [LANL]
	Contact: Robb Habbersett (505/667-0296 or robb@big-geek.lanl.gov)

(See /host/bluemoon/usr2/idllib/user_contrib/habbersett/nn_learn.pro)