Article
Computer Science, Artificial Intelligence
Zhao Zhang, Fuzhen Zhuang, Meng Qu, Zheng-Yu Niu, Hui Xiong, Qing He
Summary: The paper proposes a general technique with the concept of LSUs to enable knowledge transfer among semantically similar entities or relations, addressing the data sparsity problem in the field of KGE and providing better representations of KGs.
Article
Telecommunications
Meryem Simsek, Oner Orhan, Marcel Nassar, Oguz Elibol, Hosein Nikopour
Summary: As cellular networks continue to grow in density, traditional fiber backhaul access to each cell site becomes difficult. Utilizing millimeter wave communication and beamforming, high-speed wireless backhaul can be achieved. Our proposed topology formation approach, based on deep reinforcement learning and graph embedding, offers a less complex and more scalable solution with significant performance gains compared to baseline approaches.
IEEE COMMUNICATIONS LETTERS
(2021)
Article
Mathematical & Computational Biology
Xiaodong Zhu, Liehui Jiang, Zeng Chen
Summary: The proposed method in this paper, based on NMT and graph embedding, shows improvements in accuracy and efficiency compared to traditional methods and existing deep-learning methods.
MATHEMATICAL BIOSCIENCES AND ENGINEERING
(2021)
Article
Engineering, Civil
Shaojun Zhu, Makoto Ohsaki, Kazuki Hayashi, Shaohan Zong, Xiaonong Guo
Summary: This paper proposes a framework for critical element identification and demolition planning of frame structures using reinforcement learning and graph embedding. Through numerical examples, it is demonstrated that the trained agent can accurately estimate Q values and handle problems with different action spaces. By comparing with conventional methods, it is shown that the reinforcement learning model reduces computational cost and Q values can be used as quantitative indices for decision-making.
FRONTIERS OF STRUCTURAL AND CIVIL ENGINEERING
(2022)
Article
Computer Science, Artificial Intelligence
Kazuki Hayashi, Makoto Ohsaki
Summary: A combined method of graph embedding and reinforcement learning is developed for discrete cross-section optimization of planar steel frames. The proposed method uses an improved graph embedding formulation to extract edge features associated with members. The trained agent estimates accurate returns for each action and adjusts the size of members accordingly.
ADVANCED ENGINEERING INFORMATICS
(2022)
Article
Engineering, Industrial
Pai Zheng, Liqiao Xia, Chengxi Li, Xinyu Li, Bufan Liu
Summary: This study introduces a multi-agent reinforcement learning method based on industrial knowledge graph to achieve a Self-X cognitive manufacturing network. By establishing IKG, performing embedding algorithm, and implementing a decentralized system, it enables self-configurable solution searching, task decomposition, and self-optimization.
JOURNAL OF MANUFACTURING SYSTEMS
(2021)
Article
Computer Science, Artificial Intelligence
Guanting Du, Fei Zhu, Quan Liu
Summary: This paper proposes a critical node search algorithm (VNS) based on deep reinforcement learning, which can quickly and accurately find critical nodes in a network by utilizing graph embedding and deep Q network methods. The VNS method outperforms other algorithms in terms of both time complexity and performance, and demonstrates strong generalization performance.
CONNECTION SCIENCE
(2022)
Article
Computer Science, Information Systems
Chenghao Xu, Wei Song
Summary: Mobile crowdsensing (MCS) leverages smart device owners to cooperate with network infrastructures, combining machine intelligence and human intelligence. Data uploading is a challenging problem in MCS, and a deep reinforcement learning (DRL)-based method is proposed to address the limitations of current approaches.
IEEE INTERNET OF THINGS JOURNAL
(2022)
Article
Computer Science, Information Systems
Yandi Li, Haobo Gao, Yunxuan Gao, Jianxiong Guo, Weili Wu
Summary: The article introduces the importance and challenges of Influence Maximization (IM) problem, compares traditional algorithms with ML-based methods, and summarizes the recent research progress. It emphasizes the advantages and application prospects of using machine learning methods such as Deep Reinforcement Learning to solve the IM problem, and points out the challenges that need to be addressed in future research.
ACM TRANSACTIONS ON KNOWLEDGE DISCOVERY FROM DATA
(2023)
Article
Engineering, Civil
Kazuki Hayashi, Makoto Ohsaki, Masaya Kotera
Summary: This study considers a truss as a graph and combines graph embedding and reinforcement learning to develop an agent that can generate stable assembly paths for trusses with arbitrary configurations. The trained agent significantly outperforms metaheuristic approaches in terms of speed and applicability.
JOURNAL OF THE INTERNATIONAL ASSOCIATION FOR SHELL AND SPATIAL STRUCTURES
(2022)
Article
Computer Science, Information Systems
Qian Liu, Jie Lu, Guangquan Zhang, Tao Shen, Zhihan Zhang, Heyan Huang
Summary: The paper proposes an unsupervised meta-embedding method that dynamically aggregates embeddings from multiple sources and enriches domain-specific knowledge, providing more accurate word meta-embeddings for NLP tasks in the task domain.
INFORMATION SCIENCES
(2021)
Article
Computer Science, Information Systems
Zulie Pan, Taiyan Wang, Lu Yu, Yintong Yan
Summary: This paper presents PDM, a graph-based method that improves the accuracy of binary function similarity detection by considering position distribution information. The experiments show that PDM outperforms other tools in terms of accuracy and detection results.
Article
Automation & Control Systems
Yuwei Li, Shouling Ji, Chenyang Lyu, Yuan Chen, Jianhai Chen, Qinchen Gu, Chunming Wu, Raheem Beyah
Summary: This article introduces an evolutionary fuzzing framework called V-Fuzz, which efficiently identifies vulnerabilities in binary programs using a vulnerability prediction model and an evolutionary algorithm. Experimental results demonstrate that V-Fuzz is effective in finding bugs and has discovered new vulnerabilities.
IEEE TRANSACTIONS ON CYBERNETICS
(2022)
Article
Computer Science, Information Systems
Sarah G. Elnaggar, Ibrahim E. Elsemman, Taysir Hassan A. Soliman
Summary: Graph classification is an important task in graph data analysis, as graphs are used to illustrate relationships between entity pairs. Graphs are essential in various domains, and finding an effective way to represent their structures is crucial. This paper investigates different graph embedding methods inspired by NLP's Doc2Vec model, to represent graphs as fixed-length feature vectors for classification. Two supervised classifiers, a deep neural network and a convolutional neural network, are proposed and achieve competitive results on benchmark datasets, outperforming traditional methods and deep-learning-based approaches on three out of five datasets, with an impressive accuracy rate of 94% on the NCI1 dataset.
Article
Chemistry, Multidisciplinary
Xiaochun Sun, Chenmou Wu, Shuqun Yang
Summary: With the proliferation of Knowledge Graphs (KGs), knowledge graph completion (KGC) has attracted much attention. Previous methods focused on extracting shallow structural information from KGs or combining with external knowledge. To address the limitations, a novel Scalable Formal Concept-driven Architecture (SFCA) was proposed to encode factual triples into formal concepts, providing valuable information for KGC. Comprehensive experiments on public datasets and industry dataset demonstrated the effectiveness and scalability of SFCA, offering new ideas for the promotion and application of knowledge graphs in AI downstream tasks.
APPLIED SCIENCES-BASEL
(2023)
Article
Engineering, Civil
Zulin Huang, Hongjun Liu, Jingyao Zhang, Zhengliang Li, Makoto Ohsaki, Qiang Bai
Summary: Based on experimental tests and numerical simulations, a new design formula for the effective length coefficient of cross bracing in transmission towers has been proposed. The formula improves the consistency and reliability of predicting the ultimate strength, thus enhancing the accuracy of design calculations according to the Chinese design code.
THIN-WALLED STRUCTURES
(2023)
Article
Construction & Building Technology
Qiang Zeng, Shaojun Zhu, Zhengning Li, Xiaonong Guo
Summary: This paper presents an automatic triangular mesh-generation framework for free-form SLRSs. The algorithm utilizes non-linear virtual interaction forces (VIFs) to update joint coordinates. A mesh-energy function is defined to adjust the number and distribution of joints based on desired size. Additionally, a connectivity optimization algorithm and a mesh-smoothing method are proposed to convert the unstructured mesh into a structured type. Case studies show that this framework is robust to different types of free-form SLRSs, joint densities, sizes, and edge-swapping criteria.
AUTOMATION IN CONSTRUCTION
(2023)
Article
Engineering, Civil
Guo-Qiang Li, Jinyu Li, Shaojun Zhu
Summary: This paper presents an approach for early warning of the fire-induced collapse of planar steel trapezoid trusses based on real-time measurement of displacements and displacement velocities of key nodes. The approach is validated through a full-scale fire test and numerical simulations. By exploring the variation laws of displacements and displacement velocities, early-warning characteristic points are determined and a three-level early-warning method is proposed. The accuracy of the method is validated through fire tests and a numerical case study.
FIRE SAFETY JOURNAL
(2023)
Article
Engineering, Multidisciplinary
Wei Ji, Shaojun Zhu, Guo-Qiang Li, Bin Chen
Summary: Unexpected collapse accidents of burning buildings have posed great threats to firefighters and trapped individuals. There is an urgent need for accurate real-time early warning of fire-induced collapse of burning buildings to minimize casualties. However, current methods of early warning evaluation rely on displacement analysis, which is difficult to obtain rapidly and accurately at fire rescue scenes.
Article
Construction & Building Technology
Qiang Zeng, Xiaonong Guo, Xu Yang, Shaojun Zhu, Zhengning Li
Summary: This paper proposes a simple and efficient method to generate stochastic initial geometric imperfections (IGIs) considering topology constraints for single-layer reticulated shells (SLRSs). Numerical experiments show that the method generates IGIs that satisfy the topology constraint and result in more significant out-of-plane joint deviations. Compared to the traditional method, the proposed method has better potential to search for the lower boundary of the nonlinear buckling load of SLRSs.
JOURNAL OF STRUCTURAL ENGINEERING
(2023)
Article
Engineering, Civil
Shaojun Zhu, Zhangjianing Cheng, Chaozhong Zhang, Xiaonong Guo
Summary: In this study, a numerical analysis was conducted on aluminum alloy reticulated shells (AARSs) with gusset joints under fire conditions. The proposed thermal-structural coupled analysis model of AARSs considering joint semi-rigidity was validated and can be used for other structures. The study explored the buckling behavior of K6 AARS with gusset joints under fire conditions, identifying the factors influencing the reduction factor of the buckling capacity.
FRONTIERS OF STRUCTURAL AND CIVIL ENGINEERING
(2023)
Article
Engineering, Civil
Lei Huang, Yong Du, Shaojun Zhu, Li Wang
Summary: This study conducted uniaxial compression tests on C120 hybrid fiber ultra-high performance concrete (HF-UHPC) at elevated temperatures using a purpose-built rigid element-assisted loading system. The test results showed that C120 HF-UHPC exhibited brittle failure characteristics at temperatures between 100 degrees C to 400 degrees C, while plastic damage characteristics were observed at room temperature and between 400 degrees C to 800 degrees C. The compressive strength of C120 HF-UHPC increased with temperature from 20 degrees C to 200 degrees C and decreased within the range of 200 degrees C to 800 degrees C. Additionally, a constitutive model for C120 HF-UHPC at elevated temperatures was proposed and found to be more accurate and reasonable compared to existing models. The importance of these findings lies in providing a foundation for structural analysis of HF-UHPC structures under fire conditions.
Article
Automation & Control Systems
Jinyu Li, Guo-Qiang Li, Shaojun Zhu
Summary: This paper presents a novel framework for early warning the collapse of large-span steel truss structures in fire, which can assess the collapse state of the building in real time and improve prediction accuracy using dynamic weighted loss function and transfer learning.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Engineering, Civil
Qiang Zeng, Makoto Ohsaki, Jingyao Zhang, Shaojun Zhu, Zhengning Li, Xiaonong Guo
Summary: This paper proposes a fast and automatic method to generate structured triangular meshes for free-form SLRSs, which has been successfully applied to various types of free-form SLRSs.
ENGINEERING STRUCTURES
(2023)
Article
Engineering, Mechanical
Ryo Watada, Makoto Ohsaki
Summary: This article proposes a systematic method for designing deployable structures using hinge joints with inclined axis, which allows for a rich variety of deformations. The proposed structure, called N-gonal multilayer symmetric revolute linkage (N-MLSRL), consists of multiple layers that can be deformed from a regular frame into a straight rod shape. For N less than or equal to 3, the structure has a single degree-of-freedom. The method is applied to various numerical examples, such as a horn-shaped structure, a ball-shaped structure, and a dome-shaped structure, to validate its effectiveness.
JOURNAL OF MECHANICAL DESIGN
(2023)
Article
Engineering, Civil
Chaozhong Zhang, Shaojun Zhu, Shaohan Zong, Zhengang Sui, Xiaonong Guo
Summary: A novel assembled self-centering buckling-restrained brace (ASC-BRB) is proposed in this paper to replace damaged components and prevent weak-story mechanism of frame structure after strong earthquakes. The ASC-BRB consists of a buckling-restrained brace (BRB) system and a self-centering (SC) system. The SC system, composed of two sets of steel strands and disc springs, achieves large deformability and post-hardening behavior. The experimental results demonstrate that the ASC-BRB has excellent self-centering capability, stable energy dissipation capability, and significant deformation capacity. The refined numerical model of ASC-BRB effectively analyzes the mechanical behavior of the components and an improved configuration is proposed to enhance the stability of the system.
THIN-WALLED STRUCTURES
(2023)
Article
Engineering, Civil
Xiaonong Guo, Jindong Zhang, Qiang Zeng, Shaojun Zhu, Shaohan Zong
Summary: Structural health monitoring allows for early detection of structural damage and insights into performance. However, designing sensor placement for Single-Layer Reticulated Shells is challenging due to their complex modal characteristics. This paper proposes a novel performance-based optimal sensor placement method, using modal observability and damage identifiability as evaluation indexes. The method utilizes multi-objective optimization and performance-based design concepts, generating optimal sensor placement schemes for various performance targets. Results show its superiority over other methods and provide guidance for practical engineering applications.
THIN-WALLED STRUCTURES
(2023)
Article
Construction & Building Technology
Xiaonong Guo, Jindong Zhang, Shaohan Zong, Shaojun Zhu
Summary: This paper proposes a fast-response-generation method for SLRSs based on an implicit parameter model of generative adversarial networks (GANs). The frequency-domain response data is calculated from the time-domain response data obtained from monitoring using Fourier transform, and an implicit parameter model based on frequency-domain response of SLRS is proposed. The GAN dataset is then constructed based on the frequency-domain response, and the corresponding implicit parameter model is built. The validity of the proposed implicit parameter model is verified through numerical and application examples.
JOURNAL OF BUILDING ENGINEERING
(2023)
Article
Architecture
Jingyao Zhang, Makoto Ohsaki
Summary: This paper presents an approach for designing globally developable architectural surfaces, which have zero Gaussian curvature at every interior node. The approach breaks down the highly non-linear design problem into two sub-problems: finding member lengths of a triangular mesh with zero Gaussian curvature, and optimizing the final geometry subject to boundary and length constraints. Numerical examples demonstrate the efficiency and accuracy of solving these sub-problems. The paper also introduces an improved circle packing scheme for better conformality in the Ricci flow algorithm.
JAPAN ARCHITECTURAL REVIEW
(2023)