class TestLshLookupTable(TestCase):
hash = hdhash.HdHash(WINDOW_SIZE / BAND * NR_BIT_PER_WORD, 8)
l = lsh.Lsh(WINDOW_SIZE, BAND, ROW, hash)
def test_one_signature(self):
#given
signature = [0x0]
#when
lookup = self.l.lookup_table(signature)
#then
self.assertEqual(1, len(lookup))
self.assertEqual(['s0b0', [-1]], lookup[0])
return
def test_n_signature(self):
#given
signature = [0xff, 0]
#when
lookup = self.l.lookup_table(signature)
#then
self.assertEqual(2, len(lookup))
return
def test_n_band_signature(self):
#given
nr_unique_signature = 9
signature = list(range(nr_unique_signature)) * BAND
#when
lookup = self.l.lookup_table(signature)
#then
self.assertEqual(nr_unique_signature * BAND, len(lookup))
def test_lookup_table_with_None(self):
#given
signature = None
#when
lookup = self.l.lookup_table(signature)
#then
self.assertEqual([], lookup)
return
def test_lookup_tabke_with_empty(self):
#given
signature = []
#when
lookup = self.l.lookup_table(signature)
#then
self.assertEqual([], lookup)
def test_index_retrieval(feat_dim=10, sig_dim=100):
'''Test lsh index and retrieval'''
sim = lsh.JaccardSimilarity()
lshash = lsh.Lsh(feat_dim, sig_dim, sim)
for _ in xrange(100):
lshash.index(lsh.LshObject(gen_random_bit_vector(feat_dim)))
for _ in xrange(5):
obj = lsh.LshObject(gen_random_bit_vector(feat_dim))
results = lshash.retrieve(obj)
print "Data retrieved for %s:" % obj
for result in results:
print result
class TestLshHashing(TestCase):
hash = hdhash.HdHash(WINDOW_SIZE / BAND * NR_BIT_PER_WORD, 8)
l = lsh.Lsh(WINDOW_SIZE, BAND, ROW, hash)
def test_hashing(self):
#given
data = [0] * WINDOW_SIZE * 2
#when
value = self.l.hashing(data)
#then
self.assertEqual(BAND * 2, len(value))
self.assertEqual('0' * BAND * 2, "".join(str(x) for x in value))
def test_random_projection(feat_dim=5, sig_dim=1000):
''' test lsh for cosine similarity'''
sim = lsh.CosineSimilarity()
lshash = lsh.Lsh(feat_dim, sig_dim, sim)
x = lsh.LshObject(gen_random_vector(feat_dim))
y = lsh.LshObject(gen_random_vector(feat_dim))
lshash.generate_signature(x)
lshash.generate_signature(y)
print x.feature
print y.feature
print "cosine similarity: " + \
str(sim.compute_similarity(x.feature, y.feature))
print "approximated similarity: " + \
str(sim.approximate_similarity(x.signature, y.signature))
def test_minhash(feat_dim=5, sig_dim=1000):
''' test lsh for jaccard similarity'''
sim = lsh.JaccardSimilarity()
lshash = lsh.Lsh(feat_dim, sig_dim, sim)
x = lsh.LshObject(gen_random_bit_vector(feat_dim))
y = lsh.LshObject(gen_random_bit_vector(feat_dim))
lshash.generate_signature(x)
lshash.generate_signature(y)
print x.feature
print y.feature
print "jaccard similarity: " +\
str(sim.compute_similarity(x.feature, y.feature))
print "approximated similarity: " +\
str(sim.approximate_similarity(x.signature, y.signature))
def __init__ (self, ss, window_size, band, row, hash):
self._ss = ss
self._window_size = window_size
self._band = band
self._row = row
self._lsh = lsh.Lsh(window_size, band, row, hash)