检索结果-内蒙古大学图书馆

Frontiers of Robotics and Software engineering (FRSE), International Conference on

作者： Lei Chen Yang Wang Yun Fei Zheng Zhen Yang Tieyong Cao Institute of Command-and-Control Engineering The Army Engineering University of PLA Nanjing China The Army Artillery and Denfense Academy of PLA Institute of Command-and-Control Engineering The Army Engineering University of PLA Nanjing China

Chinese text classification aimed to set pre-defined labels to given documents, and it is a very important research field. To deal with this problem, a novel subcharacter-enhanced hybrid network model called SHN is proposed. The proposed SHN can hierarchically extract the text semantic features. Firstly, our model decomposes Chinese words by Cangjie-code into subcharacters. Secondly, both word embeddings and subcharacters n-grams vectors are used as the input part of SHN, which is aimed to fuse both internal structure information and semantic information. Thirdly, SHN makes use of a bidirectional GRU module that contains word-level attention, to obtains semantic information from the sentences in document. Finally, a dynamic convolution neural network module is used by SHN, which aimed to extract high level features. Experiments carried out on two datasets show that SHN can obtain better results than conventional methods and is effective.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Prioritization Method of Test Data for Intelligent Software based on Multi-objective Optimization 9

Prioritization Method of Test Data for Intelligent Software ...

引用

9th International Conference on Dependable Systems and Their Applications, DSA 2022

作者： Xia, Chunyan Huang, Yifan Huang, Song Zheng, Changyou Wang, Yutong Wang, Yu College of Command and Control Engineering Army Engineering University of Pla China College of Computer and Information Technology Mudanjiang Normal University China Illinois Institute of Technology United States

ISBN: (数字)9781665488778

ISBN: (纸本)9781665488778

With the rapid development of the information industry, intelligent software testing has become one of the hot research. This paper studies how to extract useful data from the original test set to test the modified model for the intelligent software. First, the initial test data are obtained according to the label classification of the original structured documents. Second, the non-dominated sorting genetic algorithm is used to prioritize the test data. Finally, the experimental results of intelligent software show that compared with the random method, the accuracy of the penalty_money prediction model is reduced by 0.04% 34.91 %, and the accuracy of the penalty_imprisonment prediction model is reduced by 0.08% 32.63 %, indicating that the proposed method can find more defects of the penalty prediction model earlier, to reduce the cost of intelligent software testing. © 2022 IEEE.

关键词： Multiobjective optimization

来源：评论

学校读者我要写书评

暂无评论

Diving into Text Representation Learning with Deep Hashing

Diving into Text Representation Learning with Deep Hashing

引用

2021 International Conference on Electronic Information engineering and Computer Science, EIECS 2021

作者： Zou, Ao Hao, Wenning Jin, Dawei Tian, Yuan College of Command Control Engineering Army Engineering University of PLA Nanjing China

ISBN: (纸本)9781665416740

With the development of deep neural network, especially the emergence of large scale pre-trained language model, methods based on deep learning have substantially advanced the state of the art across a variety of subtasks in natural language processing. Nevertheless, in order to get better performance, the increasing model parameters and exponential growth of training time become problems that can not be ignored. In this paper, inspired by the application of deep hashing in large-scale image retrieval, text representation learning based on deep hashing technology has been extensively explored. And we use three common subtasks in natural language processing to evaluate the proposed method. Experimental results show that deep hashing can greatly reduce the physical space cost by text representation under the condition of limited performance loss, which is of great significance in large-scale text representation, and is worthy of further exploration. © 2021 IEEE.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Investigation of a semi-empirical load model of natural gas explosion in vented spaces

引用

Journal of Safety Science and Resilience 2021年第3期2卷 157-171页

作者： Shigang Yang Jiongwei Cai Ya Yang Qin Fang Qi Bao Senpei Wang National Defense College of Engineering Army Engineering University of PLANanjing 210007China 73021 Troop of PLA Hangzhou 310012China Naval Command College of PLA Nanjing 210016China

Due to the influence of many factors,the overpressure-time history load model of vented gas explosions is difficult to describe and is not conducive to further structural *** on vented gas explosion test data,this paper obtains three typical overpressure-time history curves and puts forward a new semi-empirical model-double hump model that considers gas concentration and venting pressure,and gives a formula for peak pressure and overpressure-time history *** scientificity of the model is then verified by the total impulse in the *** model is able directly reflect the load characteristics,provide reference for calculating key parameters of a vented gas explosion and provide information on the structural response under the *** model thereby has the potential to help reduce the impact of gas explosion disasters.

关键词： Vented gas explosion Load model Overpressure-time history Gas concentration Venting pressure

来源：评论

学校读者我要写书评

暂无评论

Area control of Unmanned Aerial Vehicles Swarm Based on Lanner-Jones Potential 5

Area Control of Unmanned Aerial Vehicles Swarm Based on Lann...

引用

5th International Symposium on Autonomous Systems, ISAS 2022

作者： Liu, Jintao He, Ming Tian, Wei Liu, Qiang Yu, Minggang Xu, Peng People's Liberation Army Engineering University Command and Control Engineering College Nanjing China

ISBN: (纸本)9781665487085

Unmanned Aerial Vehicle (UAV) swarm area refers to the space occupied by a certain number of UAVs. By regulating the area of UAV swarm, the reflection area of radar is affected to reduce the exposure risk. To realize the effective modulation of UAV swarm area, inspired by the gas expansion phenomenon, an area modulation method of UAV swarm based on Lanner-Jones interatomic potential model is proposed. Firstly, the area measurement method based on the minimum convex hull algorithm is proposed to solve the bottleneck that the swarm distribution cannot be quantitatively analyzed. Secondly, on the basis of Vicsek flocking interaction model, the relative distance control method based on Lanner-Jones two-body interaction potential is proposed, which can realize the effective control of the global aera with the distance between two bodies as the unique variable. Finally, the simulation verification is carried out and the effects of the two control parameters named repulsion -distance and perception-distance on swarm aera is analyzed. The conclusion is that repulsion-distance is the dominant variable of swarm aera. © 2022 IEEE.

关键词： Unmanned aerial vehicles (UAV)

来源：评论

学校读者我要写书评

暂无评论

Large Language Model-Based planning Agent with Generative Memory Strengthens Performance in Textualized World

SSRN

引用

SSRN 2024年

作者： Liu, Junyang Hao, Wenning Cheng, Kai Jin, Dawei Command and Control Engineering College Army Engineering University of PLA No.1 Haifu Xiang Jiangsu Nanjing210007 China

Artificial Intelligence is increasingly being integrated into instruction-driven task planning within various external environments, such as games, compilers, and Application Programming Interfaces. Obtaining planning scenarios from Large Language Model(LLM) greatly mitigates the time and computational costs associated with rule-based and network-based approaches, but still suffers from limited contextual horizons and a lack of real-time cognitive and interactive capabilities with real environments. The engineering application point of this paper is: To address the challenges, we introduce a large language model-based task planning agent to facilitate human-computer interaction and application of LLMs in real-world task planning activities, which has a generative memory (GM-Agent) that responds to the textualized reality of planning events. The contribution in GM-Agent leverages the inherent rules and worldviews extracted from textual data, employing generative memory modules to assimilate planning knowledge from expert experiences. This framework has demonstrated state-of-the-art performance in four out of five tasks set by LandWar, matching the performance of the Reinforcement Learning with Human Feedback(RLHF) model in knowledge quiz tasks, despite having an equivalent number of parameters. Additionally, we have performed comprehensive comparison and ablation studies involving LLMs and various configurations, providing a detailed analysis of performance influencers and highlighting areas requiring further exploration. © 2024, The Authors. All rights reserved.

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Traffic Allocation Strategy Based on Reinforcement Learning in Two-layer Satellite Network

Traffic Allocation Strategy Based on Reinforcement Learning ...

引用

2021 International Conference on Internet, Education and Information Technology, IEIT 2021

作者： Dun, Congying Tan, Shihan Jin, Fenglin Xu, Kun School of Command and Control Engineering College Army Engineering University of PLA Nanjing China

ISBN: (纸本)9781665425636

In MEO-LEO satellite network, a MEO can establish connections with multiple LEOs at the same time. In order to maximize the use of satellite network resources, traffic allocation between MEO and LEOs is very important. This paper first describes the two-layer satellite network system model and point out the importance of traffic allocation among multiple LEO to MEO. Then, by quantifying the traffic resources that can be received by MEO and traffic requirements from multiple LEOs to MEO, and taking into account the remaining visible of MEO to LEOs, a traffic allocation strategy based on Q-learning in two-layer satellite networks was proposed, which aims to maximize the utility of LEOs to MEO in a given time. Then the paper grades the LEO traffic requirements to reduce the computational and storage complexity of the system. Finally, the effectiveness of the proposed strategy is verified by simulation. © 2021 IEEE.

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Practice and Thinking of 'Data Structure and Programming' Based on Blended Teaching 11

Practice and Thinking of 'Data Structure and Programming' Ba...

引用

11th International Conference on Information Technology in Medicine and Education, ITME 2021

作者： Yanqin, Tang Weiwei, Chen Yongfen, Wu En, Yuan Lei, Shi Wenyu, Zhang Army Engineering University of Pla Command Control Engineering College Nanjing Jiangsu China

ISBN: (纸本)9781665406796

In order to improve the programming ability of students, teachers are actively seeking various new methods for research and practice. Based on the 'Data Structure and Program Design' course, we have carried out the exploration and practice of blended teaching, formulated instructional design based on OBE theory, constructed the teaching mode of online preview before class, offline class + online test in class, and online test after class. Practice shows that these measures improve the students' ability to analyze, express and solve problems. © 2021 IEEE.

关键词： Data structures

来源：评论

学校读者我要写书评

暂无评论

UAV control in Smart City Based on Space-Air-Ground Integrated Network

UAV Control in Smart City Based on Space-Air-Ground Integrat...

引用

2021 International Conference on Internet, Education and Information Technology, IEIT 2021

作者： Tan, Shihan Dun, Congying Jin, Fenglin Xu, Kun School of Command and Control Engineering College Army Engineering University of PLA Nanjing China

ISBN: (纸本)9781665425636

Unmanned Aerial Vehicle (UAV) is an important part of the wireless network system of the future smart city. As a difficult point in the large-scale application of UAV, UAV control gradually attracts people's attention. Aiming at the problems of UAV control in smart city application, a near real time online learning architecture of UAV control based on the software-defined space-air-ground integrated network (SSAG) was proposed. This architecture uses the two-layer software defined network (SDN) controller architecture of SSAG framework to separate UAV control. The upper-tier SDN controller is responsible for the scheduling of UAV configuration, while the lower-tier SDN controller is responsible for regional coordination of UAV. The upper-tier SDN controller updates the tendency of network states by acquiring network states information in time interval. By simulating the network state in the next time interval, the optimal strategy of UAV scheduling of the next time interval is obtained by using the strategy iteration algorithm. Finally, an example is given to verify that the near real-time online learning architecture can accurately predict the UAV requirement, and increase the throughput of the network system compared with the traditional approach. © 2021 IEEE.

关键词： controllers

来源：评论

学校读者我要写书评

暂无评论

Inverse Reinforcement Learning Integrated Reinforcement Learning for Single Intersection Traffic Signal control 5

Inverse Reinforcement Learning Integrated Reinforcement Lear...

引用

5th International Conference on Industrial Artificial Intelligence, IAI 2023

作者： Gu, Shiyi Zhang, Tingting Zhang, Ya Southeast University School of Automation Nanjing210096 China Ministry of Education Key Laboratory of Measurement and Control of Complex Systems of Engineering Nanjing China Army Engineering University College of Command and Control Engineering Nanjing210017 China

ISBN: (纸本)9798350325294

Traffic Signal control (TSC) optimization is vital for efficient traffic flow in urban intersections. Recent research has explored the application of Reinforcement Learning (RL) agents to tackle TSC challenges. However, the selection of an appropriate reward function in RL agent training still relies on subjective judgment and experience. To address this issue, an Inverse Reinforcement Learning (IRL) integrated RL algorithm is proposed. In the IRL learning section, expert trajectory data are collected and analyzed by Relative Entropy IRL (REIRL). The latent reward of expert policy is reconstructed and utilized in RL agent training process. In the RL control section, a Double Dueling Deep Q Network is applied under a cycle control framework. As verified in simulations, the introduction of expert experience improves the performance of the RL agent to the expert level and concurrently enables robustness to expert policy noises. © 2023 IEEE.

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：