论文部分内容阅读
针对源领域和目标领域共享知识是规则、结构和逻辑等关联规则的情况,提出一种基于马尔可夫逻辑网的关联规则迁移学习方法.首先利用伪对数似然函数将源领域中马尔可夫逻辑网表示的知识迁移到目标领域中,建立两个领域之间的关联;再通过对源领域进行自诊断、结构更新和目标领域搜索新子句,来优化映射得到的结构,进而适应目标领域的学习.实验结果表明,算法成功地映射了迁移知识,提高了学习模型的精确度.
Aiming at the situation that the sharing of knowledge in the source area and the target area is rules of association, structure and logic, this paper proposes an association rules migration learning method based on Markov logic network.Firstly, using the pseudo-logarithm likelihood function, The logic represented by the logic of the network migration to the target area, the establishment of the link between the two areas; and then through the source of the field self-diagnosis, structural updates and the search of new clauses in the target area, to optimize the mapping of the resulting structure, and thus adapt to the target Domain.Experimental results show that the algorithm successfully maps migration knowledge and improves the accuracy of the learning model.