Geek Answers Handbook

Locality Sensitive Hashing - Detecting Probabilities and Values ​​for R

Thanks to those who answered my previous questions and got me so far.

I have a table of approximately 25,000 vectors, each with 48 sizes, with values ​​from 0 to 255.

I am trying to develop a Locality Sensitive Hash algorithm ( http://en.wikipedia.org/wiki/Locality-sensitive_hashing ) to search for the nearest neighboring or nearest neighboring points.

My current LSH function is this:

def lsh(vector, r = 1.0, a = None, b = None):
    if not a:
        a = [normalvariate(10, 4) for i in range(48)]
    if not b:
        b = uniform(0, r)
    hashVal = floor((sum([a[i]*vector[i] for i in range(48)]) + b)/r)
    return int(hashVal)

My questions at this point:

A:. "normalvariate (10, 4)" . , random.normalvariate(http://docs.python.org/library/random.html#random.normalvariate), "d- , ". , , .

B: wikipedia :

d (p, q) <= R, h (p) = h (q) P1

d (p, q) >= cR, h (p) = h (q) P2

R, R, " ". (http://en.wikipedia.org/wiki/Locality-sensitive_hashing#Stable_distributions)

C: (B). , R hasing . R.

D: , ?

+5
2

, "MetaOptimize" - , .
http://metaoptimize.com/qa

, .

+2

More articles:


All Articles