Markov Logic Networks(MLN) is an important model in Statistical Relational Learning.Nowadays deterministic search methods are the main methods to learn MLNs.However

the result was not compact and the algorithms can easily get into the local optima.Aiming at solving these problems

we defined the predicate template and the clause template to reduce search space and put forward a learning algorithm using Genetic Algorithm(GA)

which code is clause template

to learn MLNs's structure

using Particle Swarm Optimization(PSO) to learn MLN's weight.We gave the fitness function which makes the algorithm convergence

design the genetic operators.We testify our algorithm can get better result by theoretical analysis and experiment comparison.