基于云计算的受限玻尔兹曼机推荐算法研究

Research on Restricted Boltzmann Machines Recommendation Algorithm Based on Cloud Computing

数据的指数级增长及算法本身的复杂性使受限玻尔兹曼机面临着计算效率的问题。在详细分析受限玻尔兹曼机的基础上,将受限玻尔兹曼机与Hadoop平台的并行计算架构相结合,提出基于云平台的受限玻尔兹曼机推荐算法。该算法通过复制机制解决数据相关性问题,并将传统的受限玻尔兹曼机过程分解为若干个Hadoop任务的循环,实现并行计算。实验结果表明,与在传统平台上的实现相比,基于Hadoop并行架构的受限玻尔兹曼机推荐算法在大体量数据集的条件下可大幅提高推荐计算效率。

Coupled with the exponential expansion of the data and the high computational complexity of Restricted Bolt- zmann Machines, efficient computing of Restricted Boltzmann Machines has become an important issue. Based on the de- tailed analysis, the article introduced Hadoop platform into Restricted Boltzmann Machines, and proposed Restricted Boltzmann Machines recommendation algorithm on cloud platform. The algorithm solves the problem of data correlation with replication mechanism, and divides traditional Restricted Boltzmann Machines process into several Hadoop jobs which implements parallel computing. In the experiments, the comparative analysis between Hadoop platform implemen- tation and the previous implementation draws the conclusion that the Hadoop platform improves Restricted Boltzmann Machines computation efficiently under conditions of large data sets.