Data reduction for Boolean matrix factorization

Description

In the associated paper, we discuss a fault in a recently published method of data reduction in Boolean matrix factorization. We demonstrate experimentally that reversing the logic of the rationale yields comparable results and, moreover, that a random approach yields comparable and on average even better results as well. This data package contains folders with PDF charts that complement the findings presented in the paper.

Institutions

• Univerzita Palackeho v Olomouci Katedra informatiky

Categories

Licence