CAJ | 학술논문

协同过滤算法在推荐系统中得到了广泛应用。其中具有代表性的是基于有向图模型的矩阵分解法和基于无向图模型的受限玻尔兹曼机。文中提出了一种基于对称受限玻尔兹曼机的协同过滤算法，该算法充分考虑并利用了推荐系统中用户和物品的对称性，通过对用户和物品分别建立了一个受限玻尔兹曼机，最后，采用回归算法对二者的结果进行融合处理。实验结果显示该方法与其他相关算法相比具有明显的优越性。
협동과려산법재추천계통중득도료엄범응용。기중구유대표성적시기우유향도모형적구진분해법화기우무향도모형적수한파이자만궤。문중제출료일충기우대칭수한파이자만궤적협동과려산법，해산법충분고필병이용료추천계통중용호화물품적대칭성，통과대용호화물품분별건립료일개수한파이자만궤，최후，채용회귀산법대이자적결과진행융합처리。실험결과현시해방법여기타상관산법상비구유명현적우월성。
Collaborative filtering is widely used in recommender systems .Two classical algorithms are matrix fac-torization and restricted Boltzmann machines (RBM).In this paper, a symmetrical RBM model is proposed .We utilize the symmetry patterns of users and items , and build two RMBs based on users and items , respectively . Predictions of two RBMs via regression models are combined .Empirical results show that the model outperforms other state-of-the-art algorithms.