作者: Meng Zhang
DOI: 10.1142/S0129054117500186
关键词: Word (group theory) 、 String searching algorithm 、 Fast Fourier transform 、 Wildcard 、 Mathematics 、 Convolution 、 Wildcard character 、 Combinatorics 、 Pattern matching 、 Blossom algorithm 、 Algorithm
摘要: We give faster methods to compute discrete convolutions. assume that all the inputs are packed, is, strings packed into words such each word is with α = ⌊ ω / log2|Σ| ⌋ characters, where w length of a machine and ∑ alphabet. The output our also contains more than one element result. approach based on word-level parallelism FFT. Given two m n ( ≥ ) characters ⌈ m/α ⌉ n/α respectively, convolution them can be computed in O(nuw log muw) time, u (⌈ log|Σ|m + 2) by Experiments show method three times using standard trick. consider problem pattern matching wildcards strings. It finds occurrences text, both which may contain wildcards. By strings, we present algorithms previous O(n m)-time algorithm, text. algorithm runs muw nlog |Σ|w log1+o(1) |Σ| mlog wlog occ) occ number input. bit-parallel wildcard for long patterns.