def setUp(self):
self.algorithm = CoverTreeAlgorithm(euclidian_distance)
# Number of neighbours
self.neighbours_count = 2
class TestCoverTreeAlgorithm(unittest.TestCase):
def setUp(self):
self.algorithm = CoverTreeAlgorithm(euclidian_distance)
# Number of neighbours
self.neighbours_count = 2
def build_object_description(self, number):
features = (TestFeature(number),)
return ObjectDescription(features)
def generate_numbers_and_queries(self, numbers_cnt, queries_cnt):
# Generate 100 random numbers and create objects
numbers = [random.random() * 100000 for idx in xrange(0, numbers_cnt)]
self.objects = []
for number in numbers:
self.objects.append(self.build_object_description(number))
# Generate 100 queries
queries = [random.random() * 100000 for idx in xrange(0, queries_cnt)]
self.query_objects = []
for query in queries:
self.query_objects.append(self.build_object_description(query))
def check_result(self, query, neighbours):
""" Do the naive query and check the distance to the furthest neighbour
is the same
"""
expected_distance = euclidian_distance(query, neighbours[-1])
distances = [euclidian_distance(query, obj) for obj in self.objects]
distances = sorted(distances)
real_distance = distances[self.neighbours_count]
self.assertTrue(expected_distance, real_distance)
def test_covertree(self):
""" Tests the correctitude of the algorithm. We are generating a small
number of queries and numbers to be able to do a cross check with a
naive search
"""
self.generate_numbers_and_queries(100, 100)
# Inserting objects in covertree
start = time.time()
for obj in self.objects:
self.algorithm.insert(obj)
# Running queries
start = time.time()
results = []
for query in self.query_objects:
query_result = self.algorithm.query(query, self.neighbours_count)
results.append((query, query_result))
# Checking results
for query, neighbours in results:
self.check_result(query, neighbours)
def test_covertree_performance(self):
""" Test the performance of the covertree algorithm. We are going to
generate a high number of queries and numbers here but we are not
going to check the results anymore.
"""
self.generate_numbers_and_queries(50000, 50000)
# Inserting objects in covertree
start = time.time()
for obj in self.objects:
self.algorithm.insert(obj)
print "Inserted %d objects in %f seconds" % (len(self.objects),
time.time() - start)
# Running queries
start = time.time()
results = []
for query in self.query_objects:
query_result = self.algorithm.query(query, self.neighbours_count)
results.append((query, query_result))
print "Ran %d queries in %f seconds" % (len(self.query_objects),
time.time() - start)
