def main():
N = 200
K = 10
B = 5
# generate fake data
data_mat = np.concatenate((sgt.genAnnulusCluster(
(0.5, 0.5), 0.4, 0.3, N), sgt.genCluster((0.5, 0.5), 0.1, 100)),
axis=0)
# get distance matrix:
dist_mat = snt.getDistanceMatrix(data_mat)
# cluster it
clustering = AgglomerativeClustering(n_clusters=2,
linkage='single',
affinity='precomputed').fit(dist_mat)
#elb_obj = sst.ElbowTest(K).fit(dist_mat, linkage='single')
jump_obj = sst.JumpTest(K, B).fit(dist_mat, linkageMethod='single')
# plot it:
#spt.plotData(data_mat, labels=clustering.labels_, fname='../media/ring.singleLink1.png')
#spt.plotElbowTest(data_mat, elb_obj)
spt.plotJumpTest(data_mat, jump_obj, fname='../media/jumpRing.png')
def main():
# set program constants:
K = 10 # max number of clusters
B = 100 # number of null realisations created by gap test
N = 100 # number of points in the input data cloud
# generate clusters:
n1, n2, n3 = 20, 20, 10
n4 = N - n1 - n2 - n3
cluster_vec = [
sgt.genCluster((0.3, 0.7), 0.1, n1),
sgt.genCluster((0.1, 0.3), 0.1, n2),
sgt.genCluster((0.7, 0.5), 0.1, n3),
sgt.genCluster((0.5, 0.1), 0.1, n4)
]
# create data set:
data_mat = np.concatenate(
(cluster_vec[0], cluster_vec[1], cluster_vec[2], cluster_vec[3]),
axis=0)
data_mat = sgt.genPoisson(N) # uniform point cloud
# convert data to network:
dist_mat = snt.getDistanceMatrix(data_mat)
den_obj = sst.DendroTest(K).fit(dist_mat)
# plot the top p levels of the dendrogram
spt.plotDendroTest(data_mat, den_obj, fname='../media/dendro3.png')
def main():
# set program constants:
K = 10 # max number of clusters
B = 100 # number of null realisations created by gap test
N = 100 # number of points in the input data cloud
# generate clusters:
n1, n2 = 30, 20
n3 = N - n1 - n2
cluster_vec = [
sgt.genCluster((0.3, 0.7), 0.1, n1),
sgt.genCluster((0.1, 0.3), 0.1, n2),
sgt.genCluster((0.7, 0.5), 0.1, n3)
]
# create data set:
data_mat = np.concatenate((cluster_vec[0], cluster_vec[1], cluster_vec[2]),
axis=0)
#data_mat = sgt.genPoisson(N) # uniform point cloud
# convert data to network:
dist_mat = snt.getDistanceMatrix(data_mat)
# perform gap test
gap_obj = sst.GapTest(K, B).fit(dist_mat)
spt.plotGapTest(data_mat, gap_obj, fname='../media/gap.png')
def main():
N = 200 # number of points in each cluster
c = (0.5, 0.5) # center of cluster
# generate the data:
#data = mgt.genCigar(0.3,0.05,c,N, rot_angle=0.3*np.pi)
#data = mgt.genDisk(0.3, c, N)
#data = mgt.genAnnulus(0.1, 0.3, c, N)
#data = mgt.genRect(0.2,0.5,c,N)
#data = mgt.genCross(0.8,0.6,0.1,c,N,rot_angle=7.)
data = mgt.genS(0.6, 0.1, c, N)
#data = mgt.genEight(0.6, 0.1,c,N)
#data = mgt.genGauss(c, 0.1, 0.02, N, rot_angle=10.)
# pairwise separation matrix:
dist = snt.getDistanceMatrix(data)
# compute filter values
fil_vec = mst.filterDensity(dist, 0.1)
#fil_vec = mst.filterEccentric(dist, 3)
#fil_vec = mst.filterLinfinity(dist)
# print stats:
st_obj = mst.StatsTable(dist)
st_obj.printStatsTable()
# plot the data:
#mpt.plotPoints(data, labels=fil_vec, fname='../media/cigar.ecc.png')
#mpt.plotShapeHist(data,dist,fname='../media/cigar.shapeHist.png')
mpt.plotPointsFilterHist(data,
fil_vec,
fname='../media/s.filterHist.den.png')
def main():
# generate clusters:
data_mat = np.concatenate((sgt.genCluster(
(0.2, 0.5), 0.15, 6), sgt.genCluster(
(0.6, 0.7), 0.15, 5), sgt.genCluster((0.7, 0.2), 0.15, 3)),
axis=0)
# convert data to network:
dist_mat = snt.getDistanceMatrix(data_mat)
#GX = snt.graphFromDistMatrix(DX)
# perform clustering on the network:
#clustering = AgglomerativeClustering(n_clusters=3).fit(X) # with raw data
clustering = AgglomerativeClustering(n_clusters=3,
linkage='average',
affinity='precomputed').fit(
dist_mat) # with distance matrix
# plot it:
#spt.plotData(X, annotate=True)
#spt.plotData(X, labels=clustering.labels_)
#spt.plotData(X, labels=clustering.labels_, fName='../media/clusters.pdf')
#spt.plotGraph(GX)
#spt.plotData(X, fName="../media/pointCloud.png")
#spt.plotClustersGraph(clustering.labels_, fname='../media/clusters.graph.png')
#spt.plotPointsGraph(X)
spt.plotClustersAndGraph(data_mat,
clustering.labels_,
ffolder='../media/',
flabel=sys.argv[1])
def main():
N = 200 # number of points in each cluster
M = 3 # number of sets in the cover
K = 5 # max number of clusters in each cover member set
B = 100 # number of null realisations created each gap test
c = (0.5, 0.5) # center of cluster
# generate the data:
data = mgt.genCigar(0.3, 0.05, c, N, rot_angle=0.3 * np.pi)
#data = mgt.genDisk(0.3, c, N)
#data = mgt.genAnnulus(0.1, 0.3, c, N)
#data = mgt.genRect(0.2,0.5,c,N)
#data = mgt.genCross(0.8,0.6,0.1,c,N,rot_angle=0.)
#data = mgt.genS(0.6, 0.1,c,N)
#data = mgt.genEight(0.6, 0.1,c,N)
#data = mgt.genGauss(c, 0.1, 0.02, N, rot_angle=10.)
# pairwise separation matrix:
dist = snt.getDistanceMatrix(data)
# print stats:
st_obj = mst.StatsTable(dist)
st_obj.printStatsTable()
# compute filter values
#fil_vec = mst.filterDensity(dist, 0.1)
#fil_vec = mst.filterEccentric(dist, 3)
fil_vec = mst.filterLinfinity(dist)
eps = 0.4
# initialise mapper object:
mp = mst.Mapper(min(fil_vec), max(fil_vec), M, eps, K, B)
mp.fit(data, dist, fil_vec)
# plot the data:
#mpt.plotPoints(data, labels=fil_vec, fname='../media/cigar.ecc.png')
#mpt.plotShapeHist(data,dist,fname='../media/cigar.shapeHist.png')
mpt.plotPointsFilterHist(data,
fil_vec,
fname='../media/cigar.filterHist.lin.png')
mpt.plotGraph(mp.clusterGraph,
'../media/cigar.mapperGraph.png',
colorAtt='avfvalue')
def fit(self, dist_mat, linkageMethod='average'):
n_points = len(dist_mat[:, 0])
print(n_points)
# compute data dendrogram:
model = AgglomerativeClustering(linkage=linkageMethod,
affinity='precomputed',
distance_threshold=0,
n_clusters=None).fit(dist_mat)
self.linkage_mat = get_linkage_matrix(model)
p = self.maxclusters + 1
distances = self.linkage_mat[:, 2][:-p:-1]
self.w_log_vec = np.log10(distances)
# generate realisations:
wnull_log_mat = np.empty([self.maxclusters, self.nreals])
for b in range(self.nreals):
distnull_mat = snt.getDistanceMatrix(sgt.genPoisson(n_points))
model = AgglomerativeClustering(linkage=linkageMethod,
affinity='precomputed',
distance_threshold=0,
n_clusters=None).fit(distnull_mat)
self.null_linkage_mat = get_linkage_matrix(model)
distances = self.null_linkage_mat[:, 2][:-p:-1]
wnull_log_mat[:, b] = np.log10(distances)
self.wnull_log_average_vec = wnull_log_mat.mean(1)
self.wnull_log_err_vec = wnull_log_mat.var(1) \
*(1 + 1/self.nreals)**0.5
self.gap = self.wnull_log_average_vec - self.w_log_vec
# estimate optimal k:
for i in range(self.maxclusters - 1):
if self.gap[i] >= self.gap[i + 1] - self.wnull_log_err_vec[i + 1]:
print(f"optimal k: {i+1}")
break
return self
def main():
# generate the data:
data = np.concatenate((sgt.genCluster(
(0.2, 0.2), 0.1, 20), sgt.genCluster(
(0.6, 0.7), 0.1, 20), sgt.genCluster((0.8, 0.2), 0.1, 20)),
axis=0)
#data = sgt.genPoisson(60)
#spt.plotData(data)
# compute distance matrix:
dist = snt.getDistanceMatrix(data)
# cluster the data:
clustering = AgglomerativeClustering(linkage='average',
affinity='precomputed',
n_clusters=3).fit(dist)
# for each cluster, print cluster and shape histogram:
#spt.plotData(data, labels=clustering.labels_)
mpt.plotShapeHist(data, dist, labels=clustering.labels_)
def main():
K = 10 # max number of clusers
# fix fake data parameters:
n1, n2, n3 = 30, 20, 15 # number of points in each cluster
R1, R2, R3 = 0.3, 0.1, 0.1 # max radius of each cluster
# generate clusters:
X1 = sgt.genCluster((0.3, 0.7), R1, n1)
X2 = sgt.genCluster((0.1, 0.3), R2, n2)
X3 = sgt.genCluster((0.7, 0.5), R3, n3)
# create data set:
n = n1 + n2 + n3 # number of points in fake data set
X = np.concatenate((X1, X2, X3), axis=0)
#X = sgt.genPoisson(n) # uniform point cloud
# convert data to network:
dist_mat = snt.getDistanceMatrix(X)
# perform elbow test:
elbow_obj = sst.ElbowTest(K).fit(dist_mat)
spt.plotElbowTest(X, elbow_obj, fname='../media/elbow.png')
def main():
N = 200 # number of points in each cluster
c = (0.5, 0.5) # center of cluster
# generate the data:
#data = mgt.genCigar(0.3,0.05,c,N, rot_angle=0.3*np.pi)
data = mgt.genDisk(0.3, c, N)
#data = mgt.genAnnulus(0.1, 0.3, c, N)
#data = mgt.genRect(0.2,0.5,c,N)
#data = mgt.genCross(0.8,0.6,0.1,c,N,rot_angle=7.)
#data = mgt.genS(0.6, 0.1,c,N)
#data = mgt.genEight(0.6, 0.1,c,N)
# pairwise separation matrix:
dist = snt.getDistanceMatrix(data)
# print stats:
st_obj = mst.StatsTable(dist)
st_obj.printStatsTable()
# plot the data:
#mpt.plotPoints(data, fname='../media/cigar.png')
mpt.plotShapeHist(data, dist, fname='../media/disk.hist.png')
