An engineer raises a Bloom filter's hash count to 20

A Bloom filter uses m = 1,000,000 bits for n = 100,000 inserted elements (m/n = 10). The false-positive rate is approximately (1 - e^(-kn/m))^k for k hash functions, minimized near k* = (m/n)·ln 2 ≈ 6.93, i.e. k=7. An engineer reasons 'more hash functions means more bits checked means fewer false positives' and bumps k up to 20, without changing m or n. What happens to the false-positive rate?

15 of 46 in Data Structures & Algorithms