def ultrametricTree(distanceMatrix):
graph = nx.Graph()
for nodeId in range(len(distanceMatrix)):
graph.add_node(nodeId, {'age':0,
'size':1.0})
nodeIds = range(len(distanceMatrix))
nextNode = len(distanceMatrix)
#print('Added nodes')
while len(distanceMatrix)> 1:
i,j = findLowestElement(distanceMatrix)
nodeAge = distanceMatrix[i][j]/2.0
sizei = graph.node[nodeIds[i]]['size']
sizej = graph.node[nodeIds[j]]['size']
graph.add_node(nextNode, {'age': nodeAge, 'size': sizei + sizej})
graph.add_edge(nodeIds[i], nextNode, {'weight': nodeAge - graph.node[nodeIds[i]]['age']})
graph.add_edge(nodeIds[j], nextNode, {'weight': nodeAge - graph.node[nodeIds[j]]['age']})
for ptr in range(len(distanceMatrix)):
distanceMatrix[ptr].append( ((distanceMatrix[ptr][i]*sizei) + (distanceMatrix[ptr][j]*sizej))/(sizei+sizej))
newElement = [((distanceMatrix[i][ptr]*sizei) + (distanceMatrix[j][ptr]*sizej))/(sizei+sizej) for ptr in range(len(distanceMatrix))]
newElement.append(0)
distanceMatrix.append(newElement)
nodeIds.append(nextNode)
nextNode += 1
distanceMatrix = removeEntry(removeEntry(distanceMatrix, j),i)
nodeIds.pop(j)
nodeIds.pop(i)
#print('Processed item')
#print(distanceMatrix)
return graph
def hierarchicalClustering(nodeIDs, distanceMatrix):
"""
"""
while len(nodeIDs) > 1:
i,j = findLowestElement(distanceMatrix)
newNodeId = ' '.join([nodeIDs[i], nodeIDs[j]])
# Calculate the Davg entry and add to the Distance Matrix
for ptr in range(len(distanceMatrix)):
distanceMatrix[ptr].append( ((distanceMatrix[ptr][i]) + (distanceMatrix[ptr][j]))/2.0)
newElement = [(distanceMatrix[i][ptr] + distanceMatrix[j][ptr])/2.0 for ptr in range(len(distanceMatrix))]
newElement.append(0)
distanceMatrix.append(newElement)
nodeIDs.append(newNodeId)
del nodeIDs[j]
del nodeIDs[i]
distanceMatrix = removeEntry(removeEntry(distanceMatrix, j),i)
yield newNodeId
def hierarchicalClustering(nodeIDs, distanceMatrix):
"""
"""
while len(nodeIDs) > 1:
i, j = findLowestElement(distanceMatrix)
newNodeId = ' '.join([nodeIDs[i], nodeIDs[j]])
# Calculate the Davg entry and add to the Distance Matrix
for ptr in range(len(distanceMatrix)):
distanceMatrix[ptr].append(
((distanceMatrix[ptr][i]) + (distanceMatrix[ptr][j])) / 2.0)
newElement = [(distanceMatrix[i][ptr] + distanceMatrix[j][ptr]) / 2.0
for ptr in range(len(distanceMatrix))]
newElement.append(0)
distanceMatrix.append(newElement)
nodeIDs.append(newNodeId)
del nodeIDs[j]
del nodeIDs[i]
distanceMatrix = removeEntry(removeEntry(distanceMatrix, j), i)
yield newNodeId
def neighbourJoining(nodeIds, nextNode, distanceMatrix):
"""
"""
if len(distanceMatrix) == 2:
graph = nx.Graph()
graph.add_edge(nodeIds[0], nodeIds[1],
{'weight': distanceMatrix[0][1]})
return graph
neighbourMatrix = constructNearestNeighbourMatrix(distanceMatrix)
i, j = findMinimumElement(neighbourMatrix)
deltaLength = (sum(distanceMatrix[i]) -
sum(distanceMatrix[j])) / (len(distanceMatrix) - 2)
limbLength1 = (distanceMatrix[i][j] + deltaLength) / 2.0
limbLength2 = (distanceMatrix[i][j] - deltaLength) / 2.0
node1 = nodeIds[i]
node2 = nodeIds[j]
# Add new node to distance matrix
for m in range(len(distanceMatrix)):
distanceMatrix[m].append((distanceMatrix[m][i] + distanceMatrix[m][j] -
distanceMatrix[i][j]) / 2.0)
newRow = [
(distanceMatrix[m][i] + distanceMatrix[m][j] - distanceMatrix[i][j]) /
2.0 for m in range(len(distanceMatrix))
]
newRow.append(0)
distanceMatrix.append(newRow)
nodeIds.append(nextNode)
# Remove i & j elements
distanceMatrix = removeEntry(removeEntry(distanceMatrix, j), i)
nodeIds.pop(j)
nodeIds.pop(i)
graph = neighbourJoining(nodeIds, nextNode + 1, distanceMatrix)
graph.add_edge(node1, nextNode, {'weight': limbLength1})
graph.add_edge(node2, nextNode, {'weight': limbLength2})
return graph
def ultrametricTree(distanceMatrix):
graph = nx.Graph()
for nodeId in range(len(distanceMatrix)):
graph.add_node(nodeId, {'age': 0, 'size': 1.0})
nodeIds = range(len(distanceMatrix))
nextNode = len(distanceMatrix)
#print('Added nodes')
while len(distanceMatrix) > 1:
i, j = findLowestElement(distanceMatrix)
nodeAge = distanceMatrix[i][j] / 2.0
sizei = graph.node[nodeIds[i]]['size']
sizej = graph.node[nodeIds[j]]['size']
graph.add_node(nextNode, {'age': nodeAge, 'size': sizei + sizej})
graph.add_edge(nodeIds[i], nextNode,
{'weight': nodeAge - graph.node[nodeIds[i]]['age']})
graph.add_edge(nodeIds[j], nextNode,
{'weight': nodeAge - graph.node[nodeIds[j]]['age']})
for ptr in range(len(distanceMatrix)):
distanceMatrix[ptr].append(
((distanceMatrix[ptr][i] * sizei) +
(distanceMatrix[ptr][j] * sizej)) / (sizei + sizej))
newElement = [((distanceMatrix[i][ptr] * sizei) +
(distanceMatrix[j][ptr] * sizej)) / (sizei + sizej)
for ptr in range(len(distanceMatrix))]
newElement.append(0)
distanceMatrix.append(newElement)
nodeIds.append(nextNode)
nextNode += 1
distanceMatrix = removeEntry(removeEntry(distanceMatrix, j), i)
nodeIds.pop(j)
nodeIds.pop(i)
#print('Processed item')
#print(distanceMatrix)
return graph
def neighbourJoining(nodeIds, nextNode, distanceMatrix):
"""
"""
if len(distanceMatrix) == 2:
graph = nx.Graph()
graph.add_edge(nodeIds[0], nodeIds[1], {"weight": distanceMatrix[0][1]})
return graph
neighbourMatrix = constructNearestNeighbourMatrix(distanceMatrix)
i, j = findMinimumElement(neighbourMatrix)
deltaLength = (sum(distanceMatrix[i]) - sum(distanceMatrix[j])) / (len(distanceMatrix) - 2)
limbLength1 = (distanceMatrix[i][j] + deltaLength) / 2.0
limbLength2 = (distanceMatrix[i][j] - deltaLength) / 2.0
node1 = nodeIds[i]
node2 = nodeIds[j]
# Add new node to distance matrix
for m in range(len(distanceMatrix)):
distanceMatrix[m].append((distanceMatrix[m][i] + distanceMatrix[m][j] - distanceMatrix[i][j]) / 2.0)
newRow = [
(distanceMatrix[m][i] + distanceMatrix[m][j] - distanceMatrix[i][j]) / 2.0 for m in range(len(distanceMatrix))
]
newRow.append(0)
distanceMatrix.append(newRow)
nodeIds.append(nextNode)
# Remove i & j elements
distanceMatrix = removeEntry(removeEntry(distanceMatrix, j), i)
nodeIds.pop(j)
nodeIds.pop(i)
graph = neighbourJoining(nodeIds, nextNode + 1, distanceMatrix)
graph.add_edge(node1, nextNode, {"weight": limbLength1})
graph.add_edge(node2, nextNode, {"weight": limbLength2})
return graph
