def test_simple_kmeans(self):
# Test out a simple kmeans example.
# Suppose we have two-dimensional data points
# (1,1), (1,2), (2,2), (3,3), (5,5)
# 3 clusters
# initial means: [0, 1, 4]
dataset = [ [1,1], [1,2], [2,2], [3,3], [5,5] ]
# initial means:
# [ 1, 1 ], [ 1, 2 ], [ 5, 5 ]
# new assignment:
# [ 0, 1, 1, 1, 2 ]
# next means:
# [ 1, 1 ], [ 2, 7 / 3.0 ], [ 5, 5 ]
# new assignment:
# [ 0, 0, 1, 1, 2 ]
# next means:
# [ 1, 1.5 ], [ 2.5, 2.5 ], [ 5, 5 ]
# total error:
# (0.5^2 + 0.5^2 + 2 * 0.5^2 + 2 * 0.5^2) / 5.0
(means, error) = clust.kmeans(dataset, 3, [ 0, 1, 4 ])
print means
self.assertEqual(1, means.count([5, 5]))
self.assertEqual(1, means.count([1, 1.5]))
self.assertEqual(1, means.count([2.5, 2.5]))
self.assertAlmostEqual(6 * math.pow(0.5, 2) / 5.0, error)
def task(self):
dataset = clust.parse_input(open('adults.txt', 'r'), 1000)
all_errors = []
# Generate a random initial assignment
for clusters in self.CLUSTERS:
# get the lowest error over three runs
errors = [
clust.kmeans(dataset, clusters)[1]
for i in xrange(self.SAMPLES)
]
error = sorted(errors)[0]
all_errors.append({"x": clusters, "y": error})
chart = {
"chart": {
"defaultSeriesType": "line"
},
"xAxis": {
"title": {
"text": "Clusters"
},
"min": 1
},
"yAxis": {
"title": {
"text": "Mean Squared Error"
}
},
"title": {
"text": "K-means Results"
},
"series": [{
"data": all_errors
}]
}
return chart
def test_simple_kmeans(self):
# Test out a simple kmeans example.
# Suppose we have two-dimensional data points
# (1,1), (1,2), (2,2), (3,3), (5,5)
# 3 clusters
# initial means: [0, 1, 4]
dataset = [[1, 1], [1, 2], [2, 2], [3, 3], [5, 5]]
# initial means:
# [ 1, 1 ], [ 1, 2 ], [ 5, 5 ]
# new assignment:
# [ 0, 1, 1, 1, 2 ]
# next means:
# [ 1, 1 ], [ 2, 7 / 3.0 ], [ 5, 5 ]
# new assignment:
# [ 0, 0, 1, 1, 2 ]
# next means:
# [ 1, 1.5 ], [ 2.5, 2.5 ], [ 5, 5 ]
# total error:
# (0.5^2 + 0.5^2 + 2 * 0.5^2 + 2 * 0.5^2) / 5.0
(means, error) = clust.kmeans(dataset, 3, [0, 1, 4])
print means
self.assertEqual(1, means.count([5, 5]))
self.assertEqual(1, means.count([1, 1.5]))
self.assertEqual(1, means.count([2.5, 2.5]))
self.assertAlmostEqual(6 * math.pow(0.5, 2) / 5.0, error)
def vclust(cond, q=1, vrep='irate', vargs={}, cmeth='kmeans', nclust=16, cargs={}):
evts = dist.VREPS[vrep](cond['evts'], **vargs)
if cmeth == 'tree':
t = clust.vtree(evts, **cargs)
clsts = t.cut(nclust)
elif cmeth == 'kmeans':
clsts = clust.kmeans(evts, nclust, **cargs)
elif cmeth == 'mixmod':
clsts = clust.mixmodpartition(evts, nclust, **cargs)
#cargs should contain eg. {'model':"Gaussian_pk_Lk_Bk", reps:1}
else:
raise StandardError("don't know clustering method %s" % cmeth)
return cond.fuse(cond.new(evts=clsts))
def task(self):
dataset = clust.parse_input(open('adults.txt', 'r'), 1000)
all_errors = []
# Generate a random initial assignment
for clusters in self.CLUSTERS:
# get the lowest error over three runs
errors = [ clust.kmeans(dataset, clusters)[1] for i in xrange(self.SAMPLES) ]
error = sorted(errors)[0]
all_errors.append({ "x": clusters, "y": error })
chart = {"chart": {"defaultSeriesType": "line"},
"xAxis": {"title": {"text": "Clusters"}, "min": 1 },
"yAxis": {"title": {"text": "Mean Squared Error"}},
"title": {"text": "K-means Results"},
"series": [ {"data": all_errors } ]}
return chart
def test_kmeans(self):
for i in xrange(2, 5):
# Test kmeans with many different initial assignments.
(means, error) = clust.kmeans(self.dataset, i)
self.assertEqual(i, len(means))
def test_kmeans(self):
for i in xrange(2, 5):
# Test kmeans with many different initial assignments.
(means, error) = clust.kmeans(self.dataset, i)
self.assertEqual(i, len(means))
