Efficient Random Sampling -- Parallel, Vectorized, Cache-Efficient, and Online — mentioned by lorenzhs

Thu, 08 Jun 2017 09:08:40 GMT

The sampling algorithm you describe becomes pretty slow as k grows. There are much faster algorithms. We wrote a paper on fast sampling: https://arxiv.org/pdf/1610.05141.pdf (page 11, what you described is labelled "H" (for hash table) in the plot. Sorry about the single-letter names, it's a weird convention for sampling algorithms). You can see that as soon as k is more than a few thousand, "R" ("recursive") is much faster. What it does is recursively split the range (initially 1...N) in half, and deciding how many samples come from the left and how many from the right side. It does this by…

Paper and discussion history. A mention is not an endorsement.

Efficient Random Sampling -- Parallel, Vectorized, Cache-Efficient, and Online · HackerPapers discussions

Efficient Random Sampling -- Parallel, Vectorized, Cache-Efficient, and Online — mentioned by lorenzhs