2009-01-29: 1.20

	. Support for cluster trimming has been dropped
	. New functionality: PAM stage can now optionally respect the dendrogram. 
	  If this option is set, for each object PAM stage will only consider those
	  clusters that are on the same branch as the object being PAMed. In
	  other words, an object can only be asigned by PAM to one of the clusters lying
	  below the object on the branch on which the object is merged.
	  Intuitively, this should produce a label assignment that is more compatible
	  with the dendrogram. It will also prevent assignment of far outlying
	  objects, resulting in more unassigned objects and cleaner, tighter
	  clusters.

2008-06-12: 

	. Changed verbose output from stage 1,2 to Tree Cut, PAM stage
	. distM is only modified (copied) if necessary

2008-06-03: 1.12

	. labelUnlabeled is now deprecated and replaced by (renamed to) pamStage.
	. deepSplit range extended to 0 through 4

2008-04-03: 

	cleanup: function .assignModuleColor commented out to remove a R CMD
	check warning.

2008-03-12: 1.11-3

	another bug in interpretation of deepSplit fixed; in particular, the default now works.

2008-03-07: 1.11-2

	bug in interpretation of deepSplit fixed

2008-03-05: 1.11-1

	dynamicHybrid: When number of merges below the cut is smaller than minClusterSize, simply return
	with all labels=0 instead of throwing an error.

2008-02-19: 1.11

	change default cutHeight from max(dendro$height) to 99% of the range
	of between maximum and 5th percentile (from minimum) of dendro$height

	change in printFlush: use cat, not print to get rid of the [1] " leading the output

2008-02-16: 1.10-02

	cutreeDynamic: if method=="hybrid" and distM not given, issue a warning and default to method="tree"

2007-12-11: 1.10

	cutreeHybrid: distM is a mandatory parameter now (no more NULL
	default) since it is used in stage 1 as well.

2007-11-22: 1.09

	cutreeDynamicTree: bug where the last element on the dendrogram remained unlabeled is fixed by
	Bin Zhang.

2007-11-13: 1.09

	Changes in Dynamic Tree: minimum significant forward runlength is now
	set as a fraction (1/3)+1 of the minimum cluster size;
	minAttachModuleSize is now 2*minModuleSize instead of a hardcoded
	number 100.

2007-09-17

	Updated URL in description to http://www.genetics.ucla.edu/labs/horvath/CoexpressionNetwork/BranchCutting/


