Search

Your search keyword '"LIU Na"' showing total 394 results
Start Over Author "LIU Na" Publisher ieee
394 results on '"LIU Na"'

1. A Parallel Two-Channel Emotion Classification Method for Chinese Text

Author

Liu Na, Tao Cao, Shuchen Bai, and Danqing Li

Subjects
Attention model, Chinese text classification, deep neural networks, dual-channel model, Ernie, Electrical engineering. Electronics. Nuclear engineering, TK1-9971
Abstract

The complex structure of Chinese, including the dual-granularity feature of words and phrases, poses unique challenges for sentiment analysis, which is significantly different from the alphabetic word formation mechanism of English text. To capture Chinese semantics more accurately, an advanced parallel channel sentiment classification strategy is designed. In this study, the advanced pre-training model ERNIE with Word2vec is first adopted to enrich the embedding of Chinese text at word granularity. Following that, fine-grained features at the word level are extracted by a multi-window convolutional strategy, and the word vector sequences are deeply learnt using BiGRU network to ensure the comprehensive capture of contextual information in both directions. To further optimise the model, an Attention mechanism is introduced to ensure effective delivery of information and improve computational efficiency. After this series of innovative designs, the microblog comment dataset is used as an experimental case, and the hyperparameters are adjusted to determine the optimal parameters. Six comparison models are selected to verify the effectiveness of MsCBA on three datasets. The classification accuracies of the proposed model on the three datasets are 93.64%, 90.00% and 92.61%, respectively, which are better than the comparison models. This study provides an efficient and innovative approach for Chinese sentiment analysis, which sheds new light on the field of NLP.

Published
2024
Full Text
View/download PDF

2. An Improved Dynamic Collaborative Filtering Algorithm Based on LDA

Author

Meng Di-Fei, Liu Na, Li Ming-Xia, and Su Hao-Long

Subjects
Collaborative filtering, LDA, topic model, time tag, Electrical engineering. Electronics. Nuclear engineering, TK1-9971
Abstract

Currently, available collaborative filtering (CF) algorithms often utilize user behavior data to generate recommendations. The similarity calculation between users is mostly based on the scores, without considering the explicit attributes of the users with profiles, as these are difficult to generate, or their preferences over time evolve. This paper proposes a collaborative filtering algorithm named hybrid dynamic collaborative filtering (HDCF), which is based on the topic model. Considering that the user’s evaluation of an item will change over time, we add a time-decay function to the subject model and give its variational inference model. In the collaborative filtering score, we generate a hybrid score for similarity calculation with the topic model. The experimental results show that this algorithm has better performance than currently available algorithms on the MovieLens dataset, Netflix dataset and la.fm dataset.

Published
2021
Full Text
View/download PDF

29. Recovering Bathymetry From Satellite Altimetry-Derived Gravity by Fully Connected Deep Neural Network.

Author

Yang, Lei, Liu, Min, Liu, Na, Guo, Jinyun, Lin, Lina, Zhang, Yuyuan, Du, Xing, Xu, Yongsheng, Zhu, Chengcheng, and Wang, Yongkang

Abstract

The topography of the seafloor is highly correlated with the local gravity through intrinsically nonlinear relationships across a particular wavelength band. The purpose of this study is to compare a fully connected deep neural network (FC-DNN) and a convolutional neural network (CNN) with the gravity-geological method (GGM) to determine whether deep learning can provide superior predictions of bathymetry. We include the short-wavelength gravity (SG) and geological models as training parameters, and assess the performance of different models and parameter combinations using various inputs. Compared with the CNN method, the FC-DNN with the SG as an input reduces the standard deviation (STD) of bathymetry differences from 118.6 m to about 73.5 m. The FC-DNN with SG reduces the STD of bathymetry differences by up to 13.3% compared with the conventional GGM. Furthermore, we demonstrate that the addition of geological information alongside the SG does not significantly enhance the accuracy. Power spectral density analysis suggests that the FC-DNN is superior for predicting wavelengths shorter than 6 km. [ABSTRACT FROM AUTHOR]

Published
2023
Full Text
View/download PDF

30. Remote Sensing Image Fusion With Task-Inspired Multiscale Nonlocal-Attention Network.

Author

Liu, Na, Li, Wei, Sun, Xian, Tao, Ran, and Chanussot, Jocelyn

Abstract

Recently, convolutional neural networks (CNNs) have been developed for remote sensing image fusion (RSIF). To obtain competitive fusion performance, network design becomes more complicated by stacking convolutional layers deeper and wider. However, problems still remain when applying the existing networks in practical applications. On the one hand, researchers focus on improving spatial resolution but ignore that the fused images will be used in subsequent interpretation applications, e.g., objection detection. On the other hand, RSIF involves different tasks with different image sources, e.g., pansharpening of the panchromatic and multispectral image (MSI), hypersharpening of the panchromatic and hyperspectral image (HSI), and so on. However, the existing networks only solve one of them, failing to be compatible with other tasks. To address the above problems, a convenient task-inspired multiscale nonlocal-attention network (MNAN) is proposed for RSIF. The proposed MNAN focuses more on enhancing the multiscale targets in the scene when improving the resolution of the fused image. In addition, the proposed network can be applied to both pansharpening and hypersharpening tasks without any modification. [ABSTRACT FROM AUTHOR]

Published
2023
Full Text
View/download PDF

37. Distributed Consensus of Networked Lagrangian Systems With Unknown Nonidentical Control Directions

Author

Jianwei Zhang, Gang Wang, Qingdu Li, Song Tang, Liu Na, and Yunfeng Ji

Subjects
Consensus, General Computer Science, Computer science, General Engineering, Mode (statistics), Topology (electrical circuits), Directed graph, Revolute joint, Topology, adaptive control, Lagrangian system, uncertain dynamics, Lagrangian systems, Uniform boundedness, General Materials Science, lcsh:Electrical engineering. Electronics. Nuclear engineering, Laplacian matrix, lcsh:TK1-9971
Abstract

In the literature, most of the consensus methods require the control directions of all agents to be known. This paper deals with the distributed consensus problem without such requirements for networked Lagrangian systems subjected to uncertain dynamics. Unknown control directions are nonidentical in the sense that the control direction related to each control input in an individual Lagrangian system can be different and unknown as well as that the control directions for different Lagrangian systems are allowed to be distinct. Regarding the communication topology, the only requirement is to contain a fixed directed spanning tree. Based on the estimated consensus value, novel auxiliary sliding mode variables are constructed and applied to develop the distributed adaptive consensus control scheme. It is proved by adopting the special properties of the Laplacian matrix on directed graphs and matrix theory that global uniform boundedness of all closed-loop signals and the asymptotic consensus can be achieved. Simulations on networked two-link revolute joint arms are provided to verify the validity of the proposed scheme.

Published
2020

50. Application of Teaching Innovation Based on robotics engineering

Author

Cao Yanling, Liu Na, Han Huanqing, and Cao Shao-yong

Subjects
Class (computer programming), Engineering, business.industry, Process (engineering), Knowledge engineering, Robotics, Engineering management, Educational robotics, Engineering education, ComputingMilieux_COMPUTERSANDEDUCATION, Robot, Artificial intelligence, business, Curriculum
Abstract

As the core major of “Internet + Industrial Intelligence”, robotics engineering is an upgrade and reconstruction of traditional engineering major. The industrial robot course is the professional core course of the Robotics Engineering. It is also a comprehensive course of multi-discipline integration, which involved mechanical engineering, automatic control, computer, sensor, electronic technology, artificial intelligence and other multi-disciplinary content. Robotics Engineering is characterized by broad foundation, great difficulty, emphasis on practice, rapid development and application of new knowledge. In the process of implementation of the teaching innovation, the new concept of engineering education was applied to propose a new form of curriculum system. Taking the projects of engineering as the study objects, disassemble the knowledge points involved in industrial robots, break the course boundaries, reshape the knowledge system, draw knowledge maps and then design teaching activities. In teaching innovation, teachers extend classroom through formation of subject competition teams, promote teaching and promote learning by competition, realize the integration of “teaching, class and competition”, build a bridge between theory and practice, then complete the transformation from knowledge learning to ability training. Besides, they also keep contact with intelligent manufacturing enterprises in Zhuhai and the Bay Area to obtain real-time new developments in enterprises. Thus, the latest information was introduced into classroom. Therefore, the meaning of “production, teaching, research and application” has been deepened. According to the characteristics of the knowledge points of the course, experts were invited to make special lectures for students which can bring them with international perspective and frontier knowledge.

Published
2021

Catalog

Books, media, physical & digital resources
See catalog results