针对多层感知机(MLP)架构无法捕获会话序列上下文中的共现关系的问题,提出了一种基于图共现增强MLP的会话推荐模型GCE-MLP。首先,利用MLP架构捕获会话序列的顺序依赖关系,同时通过共现关系学习层获得序列上下文中的共现关系,并通过信息...针对多层感知机(MLP)架构无法捕获会话序列上下文中的共现关系的问题,提出了一种基于图共现增强MLP的会话推荐模型GCE-MLP。首先,利用MLP架构捕获会话序列的顺序依赖关系,同时通过共现关系学习层获得序列上下文中的共现关系,并通过信息融合模块得到会话表示;其次,设计了特定的特征选择层,旨在扩大不同关系学习层输入特征的差异性;最后,通过噪声对比任务最大化两种关系表征之间的互信息,进一步增强对会话兴趣的表征学习。在多个真实数据集上的实验结果表明GCE-MLP的推荐性能优于目前主流的模型,验证了该模型的有效性。与最优的MLP架构模型FMLP-Rec(Filter-enhanced MLP for Recommendation)相比,在Diginetica数据集上,P@20最高达到了54.08%,MRR@20最高达到了18.87%,分别提升了2.14和1.43个百分点;在Yoochoose数据集上,P@20最高达到了71.77%,MRR@20最高达到了31.78%,分别提升了0.48和1.77个百分点。展开更多
Many Bayesian learning approaches to the multi-layer perceptron （MLP） parameter optimization have been proposed such as the extended Kalman filter （EKF）. This paper uses the unscented Kalman particle filter （UPF...Many Bayesian learning approaches to the multi-layer perceptron （MLP） parameter optimization have been proposed such as the extended Kalman filter （EKF）. This paper uses the unscented Kalman particle filter （UPF） to train the MLP in a self- organizing state space （SOSS） model. This involves forming augmented state vectors consisting of all parameters （the weights of the MLP） and outputs. The UPF is used to sequentially update the true system states and high dimensional parameters that are inherent to the SOSS moder for the MLP simultaneously. Simulation results show that the new method performs better than traditional optimization methods.展开更多
文摘针对多层感知机(MLP)架构无法捕获会话序列上下文中的共现关系的问题,提出了一种基于图共现增强MLP的会话推荐模型GCE-MLP。首先,利用MLP架构捕获会话序列的顺序依赖关系,同时通过共现关系学习层获得序列上下文中的共现关系,并通过信息融合模块得到会话表示;其次,设计了特定的特征选择层,旨在扩大不同关系学习层输入特征的差异性;最后,通过噪声对比任务最大化两种关系表征之间的互信息,进一步增强对会话兴趣的表征学习。在多个真实数据集上的实验结果表明GCE-MLP的推荐性能优于目前主流的模型,验证了该模型的有效性。与最优的MLP架构模型FMLP-Rec(Filter-enhanced MLP for Recommendation)相比,在Diginetica数据集上,P@20最高达到了54.08%,MRR@20最高达到了18.87%,分别提升了2.14和1.43个百分点;在Yoochoose数据集上,P@20最高达到了71.77%,MRR@20最高达到了31.78%,分别提升了0.48和1.77个百分点。
基金supported by the National Natural Science Foundation of China(7092100160574058)+1 种基金the Key International Cooperation Programs of Hunan Provincial Science & Technology Department (2009WK2009)the General Program of Hunan Provincial Education Department(11C0023)
文摘Many Bayesian learning approaches to the multi-layer perceptron （MLP） parameter optimization have been proposed such as the extended Kalman filter （EKF）. This paper uses the unscented Kalman particle filter （UPF） to train the MLP in a self- organizing state space （SOSS） model. This involves forming augmented state vectors consisting of all parameters （the weights of the MLP） and outputs. The UPF is used to sequentially update the true system states and high dimensional parameters that are inherent to the SOSS moder for the MLP simultaneously. Simulation results show that the new method performs better than traditional optimization methods.
