def lsh2(kmer):
def permute(x):
return (2654435761 * x) % (2**32)
shingle_len = 16
shingles = []
num_shingles = len(kmer)-shingle_len+1
global global_timer
start_time = time.clock()
for i in range(num_shingles):
shingles.append(kmerutils.int_value(kmer[i:i+6]))
global_timer += time.clock() - start_time
value = min([permute(x) for x in shingles])
return value
def lsh(kmer):
# Kind of bullshitting my way through this from info here:
# http://nlp.stanford.edu/IR-book/html/htmledition/near-duplicates-and-shingling-1.html
def permute(x, seed):
random.seed(seed)
return random.randint(0, x)
# Params to do some search exploration over:
# * shingle length
# * number of seeds to use
# * could we select the shingles in a strided pattern?
shingle_len = 6 # because 2^6 == 64, 1 bit per shingle
shingles = set()
for i in range(len(kmer)-shingle_len+1):
shingles.add(kmerutils.int_value(kmer[i:i+6]))
results = []
for seed in SEEDS:
min_permuted = min([permute(x, seed) for x in shingles])
results.append(min_permuted)
value = 0
for result in results:
value *= 2
value += result % 2
return value