Fen Zhao

Fen Zhao


Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

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

Representative Publications

View All Publications
21 May 2020 in IEEE Access
Verified
Journal Article
Open access
FZFen Zhao
Fen Zhao
YLYinguo Li
Yinguo Li
XWXinheng Wang
Xinheng Wang
LBLing Bai
Ling Bai
TLTailin Liu
Tailin Liu
Capsule network is a promising model in computer vision. It has achieved excellent results on simple datasets such as MNIST, but the performance deteriorates as data becomes complicated. In order to address this issue, we propose a deep capsule network in this paper. To deepen the capsule network, we present a new tensor capsule based routing algorithm and the corresponding convolution operation. Compared to vector capsules, tensor capsules can capture more instance-level information. Together, the relevant convolution operation is beneficial for reducing the amount of parameters in the routing process. Furthermore, we propose a dropout mechanism for vectors and tensors in order to alleviate the potential overfitting problem. Finally, we also inject the multi-scale capsules of the middle layers into a multi-scale decoder to pursue more details of an image and more clear reconstructed image. Experimental results on CIFAR10, Fashion-MNIST, and SVHN demonstrate that the proposed deep tensor network can achieve very competitive performance compared to other state-of-the-art capsule networks.

2 April 2020 in IEEE Access
Verified
Journal Article
Open access
FZFen Zhao
Fen Zhao
YLYinguo Li
Yinguo Li
LBLing Bai
Ling Bai
ZTZhen Tian
Zhen Tian
XWXinheng Wang
Xinheng Wang
With the rapid growth of advanced driving assistance technology, how to implement humancar interaction humanely and personally becomes increasingly important. Existing methods typically rely on hand-crafted data to conduct training, which cannot fully extract information hidden in text. In this paper, we tackle the problem of label data acquisition difficulties and inadequate feature extraction via semisupervised learning and multi-granularity learning. To fulfill the goal of label data acquisition and feature extraction, we present a novel semi-supervised multi-granularity convolutional neural networks (CNNs)- based (SSMGCNNs-based) model for an application in human-car interaction, which consists of the twoview- embedding (TVE) module and the multi-granularity CNNs (MGCNNs) module. The TVE module learns embeddings of text regions from the unlabeled user command data set and then integrate the learned tv-embeddings into MGCNNs, so that the learned tv-embedded regions are used as an additional input into MGCNNs’ convolution layer to solve the problem of data annotation. MGCNNs can fully extract information hidden in text by multiple convolution kernels of the same convolution layer. We compared our model with some state-of-the-art machine learning models. On the car operation command data set, the simulation results demonstrated that, compared with CNNs, our method respectively improved 5.13%, 5.64%, 3.60% and 5.34% on precision, recall, F-1 and training loss.

3 September 2019 in Applied Sciences
Verified
Journal Article
Open access
PDF file
FZFen Zhao
Fen Zhao
PLPenghua Li
Penghua Li
YLYuanyuan Li
Yuanyuan Li
JHJie Hou
Jie Hou
YLYinguo Li
Yinguo Li
With the rapid developments of Internet technology, a mass of law cases is constantly occurring and needs to be dealt with in time. Automatic classification of law text is the most basic and critical process in the online law advice platform. Deep neural network-based natural language processing (DNN-NLP) is one of the most promising approaches to implement text classification. Meanwhile, as the convolutional neural network-based (CNN-based) methods developed, CNN-based text classification has already achieved impressive results. However, previous work applied amounts of manually-annotated data, which increased the labor cost and reduced the adaptability of the approach. Hence, we present a new semi-supervised model to solve the problem of data annotation. Our method learns the embedding of small text regions from unlabeled data and then integrates the learned embedding into the supervised training. More specifically, the learned embedding regions with the two-view-embedding model are used as an additional input to the CNN’s convolution layer. In addition, to implement the multi-task learning task, we propose the multi-label classification algorithm to assign multiple labels to an instance. The proposed method is evaluated experimentally subject to a law case description dataset and English standard dataset RCV1 . On Chinese data, the simulation results demonstrate that, compared with the existing methods such as linear SVM, our scheme respectively improves by 7.76%, 7.86%, 9.19%, and 2.96% the precision, recall, F-1, and Hamming loss. Analogously, the results suggest that compared to CNN, our scheme respectively improves by 4.46%, 5.76%, 5.14% and 0.87% in terms of precision, recall, F-1, and Hamming loss. It is worth mentioning that the robustness of this method makes it suitable and effective for automatic classification of law text. Furthermore, the design concept proposed is promising, which can be utilized in other real-world applications such as news classification and public opinion monitoring.