摘要
针对传统推荐模型中无法处理稀疏大数据的缺点,提出了一种基于因子分解机(Factorization Machines,FM)的推荐模型。FM模型对所有的特征分量引入了辅助向量,改进了非线性的特征向量模型方程,并采用最小交替二乘法、随机梯度下降法和马尔科夫蒙塔罗法训练该模型。理论与实验结果证明了该模型在精确度以及速度上都优于传统模型。
Based on the shortcomings can't be handled in traditional recommended models,a recommended model based on Factorization Machines(FM)is proposed.The FM model introduces an auxiliary vector for all characteristic components,improves the nonlinear characteristic vector model equation,and trains this model by using least squares method,stochastic gradient descent method and Markov Montalo method.The theoretical and experimental results show that this model is superior to the traditional model in accuracy and speed.
出处
《长春工程学院学报（自然科学版）》
2018年第2期102-104,共3页
Journal of Changchun Institute of Technology：Natural Sciences Edition
关键词
因子分解机
模型预测
目标优化
factorization machine
model prediction
objective optimization
