作为目前最成功的主流推荐方法,奇异值分解算法(SVD)将已知的海量数据建模并通过矩阵分解降维处理来得到有效信息;非负矩阵分解(NMF)则通过分解出非负矩阵元素来解释特征意义。这两种较为成功的方法均通过对显性反馈信息进行基于矩阵分...作为目前最成功的主流推荐方法,奇异值分解算法(SVD)将已知的海量数据建模并通过矩阵分解降维处理来得到有效信息;非负矩阵分解(NMF)则通过分解出非负矩阵元素来解释特征意义。这两种较为成功的方法均通过对显性反馈信息进行基于矩阵分解的处理得到用户的喜好信息来进行群体推荐。然而,仅凭用户的显性反馈信息有时无法准确反映用户的真实喜好。为解决上述问题,提出了一种针对这两种模型的改进方法,将隐性特征和基于隐性特征的群体权重计算方法融合进经典的矩阵分解算法,其中隐性特征可以完善用户的喜好信息,基于隐性特征的群体权重计算方法则根据群体的特点给予用户相应的权重,使得推荐的准确率得到提升。对该方法在KDD Cup 2012Track1中的腾讯微博数据集上进行测试,实验结果表明在该数据集上融合方法的平均绝对偏差(MAE)和准确率要优于SVD算法与NMF算法,推荐的性能有较明显的提升。展开更多
The topic of this paper is the pursuit of cultural studies focusing on cultural hegemony, introduces the notion of the dominant groups＇ power to control society. It will also raise the issue of how hegemonic classes ...The topic of this paper is the pursuit of cultural studies focusing on cultural hegemony, introduces the notion of the dominant groups＇ power to control society. It will also raise the issue of how hegemonic classes live in 1920s. The objective is to analyze, using cultural studies, Antonio Gramsci＇s Hegemony, Scott Fitzgerald＇s The Great Gatsby in order to come to some conclusions about depictions of aristocratic classes and powers in order to dominate powerless groups. Specifically, the research focuses on Jay Gatsby＇s struggles to face the hegemony of aristocratic groups, whose affluent supremacy. In the story, the new moneyed group, represented by Jay Gatsby, lives in West Egg while the aristocratic group, represented by Tom Buchanan, lives in East Egg. Tom is always the winner because he comes from the aristocratic groups, whose prestigious family. Therefore, Gatsby always loses compete against Tom no matter how hard Gatsby tries. By learning Gatsby＇s struggle in this novel, we gain a better understanding of how other powerless groups, not only in American society, but also other society in the world, who also struggle to compete with the aristocratic groups.展开更多
The DeGroot model is a classic model to study consensus of opinion in a group of individuals(agents). Consensus can be achieved under some circumstances. But when the group reach consensus with a convergent opinion va...The DeGroot model is a classic model to study consensus of opinion in a group of individuals(agents). Consensus can be achieved under some circumstances. But when the group reach consensus with a convergent opinion value which is not what we expect, how can we intervene the system and change the convergent value? In this paper a mechanism named soft control is first introduced in opinion dynamics to guide the group's opinion when the population are given and evolution rules are not allowed to change. According to the idea of soft control, one or several special agents,called shills, are added and connected to one or several normal agents in the original group. Shills act and are treated as normal agents. The authors prove that the change of convergent opinion value is decided by the initial opinion and influential value of the shill, as well as how the shill connects to normal agents. An interesting and counterintuitive phenomenon is discovered: Adding a shill with an initial opinion value which is smaller(or larger) than the original convergent opinion value dose not necessarily decrease(or increase) the convergent opinion value under some conditions. These conditions are given through mathematical analysis and they are verified by the numerical tests. The authors also find out that the convergence speed of the system varies when a shill is connected to different normal agents. Our simulations show that it is positively related to the degree of the connected normal agent in scale-free networks.展开更多
文摘作为目前最成功的主流推荐方法,奇异值分解算法(SVD)将已知的海量数据建模并通过矩阵分解降维处理来得到有效信息;非负矩阵分解(NMF)则通过分解出非负矩阵元素来解释特征意义。这两种较为成功的方法均通过对显性反馈信息进行基于矩阵分解的处理得到用户的喜好信息来进行群体推荐。然而,仅凭用户的显性反馈信息有时无法准确反映用户的真实喜好。为解决上述问题,提出了一种针对这两种模型的改进方法,将隐性特征和基于隐性特征的群体权重计算方法融合进经典的矩阵分解算法,其中隐性特征可以完善用户的喜好信息,基于隐性特征的群体权重计算方法则根据群体的特点给予用户相应的权重,使得推荐的准确率得到提升。对该方法在KDD Cup 2012Track1中的腾讯微博数据集上进行测试,实验结果表明在该数据集上融合方法的平均绝对偏差(MAE)和准确率要优于SVD算法与NMF算法,推荐的性能有较明显的提升。
文摘The topic of this paper is the pursuit of cultural studies focusing on cultural hegemony, introduces the notion of the dominant groups＇ power to control society. It will also raise the issue of how hegemonic classes live in 1920s. The objective is to analyze, using cultural studies, Antonio Gramsci＇s Hegemony, Scott Fitzgerald＇s The Great Gatsby in order to come to some conclusions about depictions of aristocratic classes and powers in order to dominate powerless groups. Specifically, the research focuses on Jay Gatsby＇s struggles to face the hegemony of aristocratic groups, whose affluent supremacy. In the story, the new moneyed group, represented by Jay Gatsby, lives in West Egg while the aristocratic group, represented by Tom Buchanan, lives in East Egg. Tom is always the winner because he comes from the aristocratic groups, whose prestigious family. Therefore, Gatsby always loses compete against Tom no matter how hard Gatsby tries. By learning Gatsby＇s struggle in this novel, we gain a better understanding of how other powerless groups, not only in American society, but also other society in the world, who also struggle to compete with the aristocratic groups.
基金supported by the National Natural Science Foundation of China under Grant No.61374168
文摘The DeGroot model is a classic model to study consensus of opinion in a group of individuals(agents). Consensus can be achieved under some circumstances. But when the group reach consensus with a convergent opinion value which is not what we expect, how can we intervene the system and change the convergent value? In this paper a mechanism named soft control is first introduced in opinion dynamics to guide the group's opinion when the population are given and evolution rules are not allowed to change. According to the idea of soft control, one or several special agents,called shills, are added and connected to one or several normal agents in the original group. Shills act and are treated as normal agents. The authors prove that the change of convergent opinion value is decided by the initial opinion and influential value of the shill, as well as how the shill connects to normal agents. An interesting and counterintuitive phenomenon is discovered: Adding a shill with an initial opinion value which is smaller(or larger) than the original convergent opinion value dose not necessarily decrease(or increase) the convergent opinion value under some conditions. These conditions are given through mathematical analysis and they are verified by the numerical tests. The authors also find out that the convergence speed of the system varies when a shill is connected to different normal agents. Our simulations show that it is positively related to the degree of the connected normal agent in scale-free networks.
