lshcos 0.0.2

A fast Python implementation of locality sensitive hashing with cosine distance.

A fast Python implementation of locality sensitive hashing with cosine distance

Installation

LSHCos depends on the following libraries:

• numpy

To install:

$ pip install lshcos

Quickstart

To create band 2 row 6 hashes for input data of 8 dimensions:

import numpy as np
from lshcos import LSHCos

### usage one:
lshcos = LSHCos(2, 6, 8)
lshcos.add([1,2,3,4,5,6,7,8],'v_1')
lshcos.add([2,3,4,5,6,7,8,9],'v_2')
lshcos.add([10,12,99,1,5,31,2,3],'v_3')
lshcos.query([1,2,3,4,5,6,7,7],2)

### usage two:
lshcos = LSHCos(2, 6, 8)
xxx = [('v_'+str(i),np.random.randn(8)) for i in range(100)]
yyy = np.random.randn(8)
for name,v in xxx:
    lshcos.add(v,name)
lshcos.query(yyy)

Main Interface

• To initialize a LSHCos instance:

LSHCos(band_num, row_num, v_dim)

parameters:

band_num:

The number of band.

row_num:

The number of row for each band.

v_dim:

The dimension of the input vector.