Lecture10

# Lecture10 - Lecture 10, Oct 21 2010 Bloom Filter Analysis...

This preview shows pages 1–3. Sign up to view the full content.

1 115 Bloom Filter Analysis After all elements are hashed, probability of specific bit in A still being zero: Probability of false positive is What is the “right” number of hash functions ( k ) ? » Two competing factors: Larger k increases probability of finding zero if the element is not in the set (good) Too large k will fill A with too many ones (p too small - bad) (1 p ) k p = ¡ 1 1 m ¢ kn = ¡¡ 1 1 m ¢ m ¢ kn m e kn m using (1 1 m ) m e 1 Lecture 10, Oct 21 2010 116 Analysis continued Rewrite false positive probability: We will try to minimize the argument of the exponent. Using rewrite the exponent as a function of p: From symmetry considerations, the minimum is at p=0.5: For this value of k, the false positive rate is k ln(1 e kn/m )= m n ln p ln(1 p ) p e kn/m (0 . 5) k =(0 . 62) m/n (1 p ) k ¡ 1 e kn/m ¢ k =exp( k ln(1 e kn/m )) k = m n ln 2

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
2 Example Consider m/n=10 If k=1, then false positive probability is
This is the end of the preview. Sign up to access the rest of the document.

## This note was uploaded on 03/08/2011 for the course CS 161 at Stanford.

### Page1 / 5

Lecture10 - Lecture 10, Oct 21 2010 Bloom Filter Analysis...

This preview shows document pages 1 - 3. Sign up to view the full document.

View Full Document
Ask a homework question - tutors are online