A Markov Logic Network Learning Algorithm From Relational Missing Data

Abstract

Markov logic network (MLN) is an important model of statistical relational learning. Learning MLN from data is important in constructing MLN. Real-world data usually contains missing data, learning MLN from missing data is more difficult than learning it from complete data, because we can't compute the exact number of the cases. We put forward a MLN learning algorithm MEM (MLN Expectation Maximization), it can learn MLN from relational missing data by expanding EM algorithm with our previous works. We define relational missing data, design initial MLN and complete algorithm for the relational missing data. Both theoretical analysis and experimental results show that MEM can effectively learn MLN from relational missing data.