English

Improved Algorithms for Exact and Approximate Boolean Matrix Decomposition

Discrete Mathematics 2015-12-29 v1

Abstract

An arbitrary m×nm\times n Boolean matrix MM can be decomposed {\em exactly} as M=UVM =U\circ V, where UU (resp. VV) is an m×km\times k (resp. k×nk\times n) Boolean matrix and \circ denotes the Boolean matrix multiplication operator. We first prove an exact formula for the Boolean matrix JJ such that M=MJTM =M\circ J^T holds, where JJ is maximal in the sense that if any 0 element in JJ is changed to a 1 then this equality no longer holds. Since minimizing kk is NP-hard, we propose two heuristic algorithms for finding suboptimal but good decomposition. We measure the performance (in minimizing kk) of our algorithms on several real datasets in comparison with other representative heuristic algorithms for Boolean matrix decomposition (BMD). The results on some popular benchmark datasets demonstrate that one of our proposed algorithms performs as well or better on most of them. Our algorithms have a number of other advantages: They are based on exact mathematical formula, which can be interpreted intuitively. They can be used for approximation as well with competitive "coverage." Last but not least, they also run very fast. Due to interpretability issues in data mining, we impose the condition, called the "column use condition," that the columns of the factor matrix UU must form a subset of the columns of MM.

Keywords

Cite

@article{arxiv.1512.08041,
  title  = {Improved Algorithms for Exact and Approximate Boolean Matrix Decomposition},
  author = {Yuan Sun and Shiwei Ye and Yi Sun and Tsunehiko Kameda},
  journal= {arXiv preprint arXiv:1512.08041},
  year   = {2015}
}

Comments

DSAA2015

R2 v1 2026-06-22T12:18:05.585Z