Fast perfect weighted resampling

We describe an algorithm for perfect weighted-random sampling of a population with time complexity O(m + n) for sampling m inputs to produce n outputs. This algorithm is an incremental improvement over standard sampling algorithms. Our sampling algorithm is parallelizable, with linear speedup. Linear-time sampling yields notable performance improvements in our motivating example of Sequential Importance Resampling for Bayesian Particle Filtering.

• Type: Conference
• Year: 2008
• Where: ICASSP
• Authors: Bart Massey
• Keywords: ICASSP 2008, Linear-time Sampling Yields, Perfect Weighted-random Sampling, Signal Processing, Standard Sampling Algorithms