基于降噪自动编码器的协同过滤推荐算法

Collaborative filtering recommendation algorithm based on noise reduction automatic encoder

协同过滤是推荐算法中运用最广也是最成功的技术之一,但是传统协同过滤的缺点也很明显.近年来基于深度学习神经网络的协同过滤技术备受关注,例如本文基于自动编码器的协同过滤推荐系统.自动编码器的训练过程较之于其他网络更短,精度也更为优秀,但是由于结构简单,更容易陷入局部最小导致过拟合.基于此,本文提出了降噪的自动编码器,最后在MovieLens 1M和MovieLens 10M数据集上进行验证.结果表明,降噪自动编码器具有更好的泛化能力.

Collaborative filtering is one of the most widely used and successful technologies in recommendation algorithms,but the shortcomings of traditional collaborative filtering are also obvious.In recent years,collaborative filtering technology based on deep learning neural network has attracted much attention,such as collaborative filtering recommendation system based on automatic encoder in this paper.Compared with other networks,the training process of automatic encoder is shorter and the accuracy is better.However,due to its simple structure,it is easier to fall into local minimum and resulting in over fitting.Considering this problem,this paper proposes an automatic encoder with noise reduction.Finally,it is verified on MovieLens 1M and MovieLens 10M datasets.The results show that the noise reduction automatic encoder has better generalization ability.