Reduce before you factorize
Description
In the associated paper, we propose a speed-up of existing Boolean matrix factorization algorithms by reducing the size of the input matrix. This data package contains folders with PDF charts complementing the findings presented in the paper, raw datasets, illustrative examples, and the source code for the utilized algorithms. Examples The file Examples.pdf contains all illustrative examples supplementing the paper. It includes examples demonstrating the employed notions as well as simulations of the proposed algorithms. S1 - Benchmark speed This folder contains charts for each benchmark dataset, comparing the runtimes of the modified algorithms with the original, unmodified counterparts. The charts show the number of factors found over time (seconds). The points where the lines stop increasing indicate the final runtimes of the algorithm. S2 - Benchmark coverage This folder contains results that demonstrate the importance of algorithm modifications for computing factorizations on the reduced data. The graphs with the prefix "Extended" correspond to the modified algorithms, while the others depict the behavior of the unmodified algorithms on the reduced matrices. Source This folder contains all the datasets described in the article (they essentially come from the website https://datasets.inf.upol.cz/). It also includes implementations of both the original and extended algorithms, as well as two experiments used to generate the results on the mentioned datasets. To run the experiments, see the respective help guides describing the parameters, e.g., $ python3 experiment_1.py —help.