1993 年 8 巻 2 号 p. 222-229
A hypothetical reasoning is an important knowledge system's framework because of its theoretical basis and its usefulness for practical problems including diagnosis, design, etc. The most crucial problem with the hypothetical reasoning is, however, its slow inference speed. In order to achieve practical inference speed, we present a fast inference method employing an approximate solution method of 0-1 integer programming. In this method, we regard all described knowledge as constraints. To narrow down the search space, we first find restricted knowledge relevant to the proof of a given goal. Then, we transform the restricted knowledge into inequations to apply 0-1 integer programming. While the computational complexity of hypothetical reasoning is NP-complete or NP-hard, this method allows mean-time inference speed below exponential order by relaxing the condition to find a quasioptimal solution rather than quasi-optimal one.
