近些年很多基于深度学习的推荐模型被提出,这些模型通过对特征的处理和改变深度网络结构来解决推荐系统数据稀疏和冷启动的问题.然而现有的方法忽略了特征与特征之间的交互对深度网络的影响,限制了模型的学习能力.为了给用户推荐更感兴...近些年很多基于深度学习的推荐模型被提出,这些模型通过对特征的处理和改变深度网络结构来解决推荐系统数据稀疏和冷启动的问题.然而现有的方法忽略了特征与特征之间的交互对深度网络的影响,限制了模型的学习能力.为了给用户推荐更感兴趣的项目和信息,本文提出了分解机深度网络(Factorization Machine Deep Network,FMN)模型.该模型将因式分解机和深度神经网络结合,首先利用因式分解机在特征之间进行交互以充分学习交叉项特征,然后利用深度网络学习高阶非线性特征.进而,分解机深度网络将特征的隐藏信息充分发掘出来并拥有高阶的非线性特征学习能力.两个真实数据集的实验表明,本文提出的模型在推荐性能上有着明显的提升.展开更多
基于逻辑回归、因式分解机、深度神经网络3种机器学习算法,提出了一种预判移动用户是否升级至高ARPU(Average Revenue Per User)套餐的方法。经业务域的用户数据验证,预测精准率达84%,召回率超50%,效果远优于传统的规则排序方法。研究...基于逻辑回归、因式分解机、深度神经网络3种机器学习算法,提出了一种预判移动用户是否升级至高ARPU(Average Revenue Per User)套餐的方法。经业务域的用户数据验证,预测精准率达84%,召回率超50%,效果远优于传统的规则排序方法。研究成果可帮助运营商更主动、更有针对性地开展营销活动,提高用户向高ARPU套餐的转化率,尤其是5G商用初期可扩展应用于挖掘5G潜力用户。展开更多
We give algorithms to factorize large integers in the duality computer. We provide three duality algorithms for factorization based on a naive factorization method, the Shor algorithm in quantum computing, and the Fer...We give algorithms to factorize large integers in the duality computer. We provide three duality algorithms for factorization based on a naive factorization method, the Shor algorithm in quantum computing, and the Fermat＇s method in classical computing. All these algorithms may be polynomial in the input size.展开更多
Proxy signature has drawn great concerns. However, there still remains a challenge to construct a provably secure and efficient proxy signature scheme. In this paper, we propose an efficient proxy signature scheme bas...Proxy signature has drawn great concerns. However, there still remains a challenge to construct a provably secure and efficient proxy signature scheme. In this paper, we propose an efficient proxy signature scheme based on factoring, and prove that it is secure in the random oracle. Furthermore, we present a new type of proxy signature, called Proxy Signature with Untrustworthy Proxy Signer, and construct a concrete scheme.展开更多
文摘近些年很多基于深度学习的推荐模型被提出,这些模型通过对特征的处理和改变深度网络结构来解决推荐系统数据稀疏和冷启动的问题.然而现有的方法忽略了特征与特征之间的交互对深度网络的影响,限制了模型的学习能力.为了给用户推荐更感兴趣的项目和信息,本文提出了分解机深度网络(Factorization Machine Deep Network,FMN)模型.该模型将因式分解机和深度神经网络结合,首先利用因式分解机在特征之间进行交互以充分学习交叉项特征,然后利用深度网络学习高阶非线性特征.进而,分解机深度网络将特征的隐藏信息充分发掘出来并拥有高阶的非线性特征学习能力.两个真实数据集的实验表明,本文提出的模型在推荐性能上有着明显的提升.
文摘基于逻辑回归、因式分解机、深度神经网络3种机器学习算法,提出了一种预判移动用户是否升级至高ARPU(Average Revenue Per User)套餐的方法。经业务域的用户数据验证,预测精准率达84%,召回率超50%,效果远优于传统的规则排序方法。研究成果可帮助运营商更主动、更有针对性地开展营销活动,提高用户向高ARPU套餐的转化率,尤其是5G商用初期可扩展应用于挖掘5G潜力用户。
基金The project supported by the 973 Program under Grant No. 2006CB921106, National Natural Science Foundation of China under Grant Nos. 10325521 and 60433050, and the Key Project 306020 and Science Research Fund of Doctoval Program of the Ministry of Education of China
文摘We give algorithms to factorize large integers in the duality computer. We provide three duality algorithms for factorization based on a naive factorization method, the Shor algorithm in quantum computing, and the Fermat＇s method in classical computing. All these algorithms may be polynomial in the input size.
基金the National Basic Research Program(973) of China (No. 2007CB31074)the National Natural Science Foundation of China (No. 90718001)
文摘Proxy signature has drawn great concerns. However, there still remains a challenge to construct a provably secure and efficient proxy signature scheme. In this paper, we propose an efficient proxy signature scheme based on factoring, and prove that it is secure in the random oracle. Furthermore, we present a new type of proxy signature, called Proxy Signature with Untrustworthy Proxy Signer, and construct a concrete scheme.
