Weina Zhang

Weina Zhang

Auto-generated profile, based on publicly available information

Weina Zhang

Communication University of China, China

Research Focus & Metrics

Academic Career Length

15Years (2012-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 March 2019 in Neurocomputing
Verified
Journal Article
Closed access
WZWeina Zhang
Weina Zhang
XZXingming Zhang
Xingming Zhang
HWHaoxiang Wang
Haoxiang Wang
DCDongpei Chen
Dongpei Chen

17 September 2019 in IEEE Access
Verified
Journal Article
Open access
WZWeina Zhang
Weina Zhang
XZXingming Zhanga
Xingming Zhanga
HWHaoxiang Wang.
Haoxiang Wang.
Factorization machines are important methods for discovering complex and evolving relationships among data entities by crafting combinatorial features automatically. Because of good data representation ability, recently Deep Neural Networks (DNNs) have been applied to factorization machines to learn high-order interactions. However, the interactions with DNNs are generated implicitly. In this paper, we propose a novel factorization machine named High-order Cross Factorization Machine (HCFM), which is an efficient and explicit high-order feature interaction factorization machine. The main component that we design for HCFM is an explicit cross network: Cross Weights Network (CWN). CWN considers both the weights of different feature combinations and interaction orders, which can capture the inherent correlations of real-word data. In CWN, cross and compression layers can learn different weights of feature interactions efficiently. Different weights of interaction orders can be learned by the weights pooling layer. CWN not only retains the important interaction features but also avoids the complex interaction computations. Comprehensive experiments are conducted on three real-world recommendation datasets to verify the validity of the model. The experiment results have demonstrated its superiority over the other state-of-art algorithms, in terms of both accuracy and space complexity.

23 June 2020 in Neurocomputing
Verified
Journal Article
Closed access
DCDongpei Chen
Dongpei Chen
XZXingming Zhang
Xingming Zhang
HWHaoxiang Wang
Haoxiang Wang
WZWeina Zhang
Weina Zhang

21 May 2021
Verified
Conference Paper
Open access
WZWeina Zhang
Weina Zhang
YZYilun Zhang
Yilun Zhang
XHXiang Lin Huang
Xiang Lin Huang
The Cox proportional hazards model is usually used in medical research for analyzing the association between the survival time of patients and one or more predictor variables. However, nonlinear, high-dimensional, low-sample size survival data cause computational challenges in survival analysis. An artificial neural network (ANN) is used for modeling nonlinear problems. A residual neural network is a specific type of neural network. In this paper, we build Cox proportional hazards model based on a residual neural network architecture, which we call it resnetCox model. The residual mapping of the resnetCox is two layers of fully connected network. The identity mapping of the resnetCox is a linear projection of inputs. The model is trained with the negative log likelihood of Cox model by Adam optimizer. It is implemented on the PyTorch in Python. We demonstrate the performance of resnetCox on BRCA, LUAD, KIRC and LUSC datasets from The Cancer Genome Atlas. We compare it with the other 3 models and experiments show that resnetCox has better performance than others in prediction accuracy and biological relevance.

1 November 2019
Verified
Conference Paper
Closed access
WZWeina Zhang
Weina Zhang
XZXingming Zhang
Xingming Zhang
HWHaoxiang Wang
Haoxiang Wang
As implicit feedback can be tracked automatically and is easy to collect, the implicit recommendation attracts more researcher attentions. However, the uncertainty of the implicit feedback meaning poses a great challenge to the implicit recommendation. Especially for the missing data, we are not sure whether the users dislike or just have not seen the items. It may lead to bias of predictions. In this paper, we propose Neural Fuzzy Inference based on User preference and Item popularity (UI-NFI) algorithm to model the missing data in implicit recommendation. First, we use fuzzy set theory to represent user preference and item popularity that get from the history interactions and side information. Furthermore, neural fuzzy inference is proposed to predict the exposure possibility of missing data. Based on the fuzzy inference model, UI-NFI and matrix factorization model perform joint learning to predict. Experimental results show that our model has better performance compared to the other implicit recommendation algorithms.

1 June 2012
Verified
Conference Paper
Closed access
WZWeina Zhang
Weina Zhang
FSFeng Shuang
Feng Shuang
Hua LiHua Li
Hua Li
In order to provide marketing support for online shop owner, this paper describes how to analyze historical data of online shop customer behaviors by data mining, and establishes logistic predictive modeling by SAS, then predicts whether customers purchase a Tablet PC online shop. Practice has proved that data mining techniques and predictive regression modeling can be effectively integrated in the application system, and the predictive modeling can predict the possibility of customers to buy goods. Also the predictive modeling has usability and availability.