def isolateClusters(selection, datasetName):
"""
Create a new data set from a selection of clusters.
@type selection: list
@var selection: The indices of the clusters to be merged.
@type datasetName: str
@var datasetName: The display name for the new data set.
"""
if (len(selection) > 0):
currFCData = DataStore.getCurrentDataSet()
if (datasetName == ""):
datasetName = currFCData.displayname
clusters = separate(currFCData.data, currFCData.getCurrentClustering())
# merge selected clusters
newData = dh.mergeData(selection, clusters)
# assign new data set to the store
newFCData = FacsData('', currFCData.labels, newData, parent=currFCData.ID)
newFCData.displayname = datasetName
newFCData.selDims = currFCData.selDims
# add basic text annotations
textAnn = {'parent': currFCData.displayname}
textAnn['events'] = len(newFCData.data)
newFCData.annotations['text'] = textAnn
DataStore.add(newFCData)
def merge(limit, ids, clusters, dist=None, minpair=None, newID=None):
"""
Take a set of clusters and iteratively merge them together based on some
distance metric until the lower limit is reached.
:@type limit: int
:@param limit: The target number of clusters.
:@type ids: list or array
:@param ids: A list containing the cluster ID of each row of data.
:@type clusters: list
:@param clusters: A list of arrays, each containing the
"""
if not isinstance(clusters, dict):
tmp = {}
for i, cluster in enumerate(clusters):
tmp[i] = cluster
clusters = tmp
if len(clusters) <= limit:
unique = np.unique(ids)
saved = dict(zip(unique, range(len(unique))))
# reassign cluster ids to a contiguous range
for i, id in enumerate(ids):
ids[i] = saved[id]
return ids
# calculate distance matrix
if (dist is None):
dist, minpair = distMatrix(clusters)
else:
dist, minpair = updateDistMatrix(clusters, dist, minpair, newID)
# combine the two most similar clusters
merged = dh.mergeData(minpair, clusters)
# merge the minpair clusters and reassign their IDs to the new
for i in minpair:
del clusters[i]
newID = np.max(ids) + 1
clusters[newID] = merged
for i, id in enumerate(ids):
if id in minpair:
ids[i] = newID
return merge(limit, ids, clusters, dist, minpair, newID)
