EN 600.639: Final Project with Samantha Siow, Michael Norris, Aaron Lerner, and Isac Lee

• numpy pip install numpy
• opencv (on mac, brew tap homebrew/science then brew install opencv)

Our code is in the fft directory:

$ python cli.py

fftmatch.naive_string_match_index(text, pattern)
fftmatch.fft_match_index_n_log_n(text, pattern)

Naive 1-D FFT-based match-index algorithm.

fftmatch.fft_match_index_n_log_m(text, pattern) Most efficient 1-D FFT-based FFT-based match-index algorithm.

boyermoore.boyer_moore_match_index(text, pattern)

Used to benchmark all of our algorithms with:

fftmatch.fft_match_index_n_sq_log_n(texts, pattern)

Similar to the n log n 1-D algorithm.

fftmatch.fft_match_index_n_sq_log_n\_naive(texts, pattern)

This uses the 1-D algorithm on each text individually from a list of texts.

fftmatch.fft_match_index_n_sq_log_m(texts, pattern)

This breaks up the text into smaller chunks of size 2*len(pattern) and does a 2-D FFT on the text.

cvmatch.cv_match_index(texts, pattern)

This uses openCV's template-matching algorithm to solve the match index problem.

cvmatch.cv_match_index_chunk(texts, pattern)

This uses openCV's template-matching algorithm on size 2*len(pattern) partitions.

Run with:

sh run_analysis.sh

Writes results to the results folder. Performs analysis on the time performance of the algorithms according to text length, as well as the time performance of the nlogm algorithm by the chunk size.