Graph Collaborative Filtering Based on Dual-Message Propagation Mechanism

2021 ◽  
pp. 1-13
Author(s):  
Haodong Liu ◽  
Bo Yang ◽  
Dongsheng Li
2021 ◽  
Vol 2021 ◽  
pp. 1-12
Author(s):  
Junjie Huang ◽  
Liang Tan ◽  
Sun Mao ◽  
Keping Yu

Blockchain is a mainstream technology in which many untrustworthy nodes work together to maintain a distributed ledger with advantages such as decentralization, traceability, and tamper-proof. The network layer communication mechanism in its architecture is the core of the networking method, message propagation, and data verification among blockchain nodes, which is the basis to ensure blockchain’s performance and key features. When blocks are propagated in peer-to-peer (P2P) networks with gossip protocol, the high propagation delay of the protocol itself reduces the propagation speed of the blocks, which is prone to the chain forking phenomenon and causes double payment attacks. To accelerate the propagation speed and reduce the fork probability, this paper proposes a blockchain network propagation mechanism based on proactive network provider participation for P2P (P4P) architecture. This mechanism first obtains the information of network topology and link status in a region based on the internet service provider (ISP), then it calculates the shortest path and link overhead of peer nodes using P4P technology, prioritizes the nodes with good local bandwidth conditions for transmission, realizes the optimization of node connections, improves the quality of service (QoS) and quality of experience (QoE) of blockchain networks, and enables blockchain nodes to exchange blocks and transactions through the secure propagation path. Simulation experiments show that the proposed propagation mechanism outperforms the original propagation mechanism of the blockchain network in terms of system overhead, rate of data success transmission, routing hops, and propagation delay.


Sensors ◽  
2020 ◽  
Vol 20 (23) ◽  
pp. 6950
Author(s):  
Rongyu Liang ◽  
Feng Liu ◽  
Jie Liu

A small fault in a large communication network may cause abrupt and large alarms, making the localization of the root cause of failure a difficult task. Traditionally, fault localization is carried out by an operator who uses alarms in alarm lists; however, fault localization process complexity needs to be addressed using more autonomous and intelligent approaches. Here, we present an overall framework that uses a message propagation mechanism of belief networks to address fault localization problems in communication networks. The proposed framework allows for knowledge storage, inference, and message transmission, and can identify a fault’s root cause in an event-driven manner to improve the automation of the fault localization process. Avoiding the computational complexity of traditional Bayesian networks, we perform fault inference in polytrees with a noisy OR-gate model (PTNORgate), which can reduce computational complexity. We also offer a solution to store parameters in a network parameter table, similar to a routing table in communication networks, with the aim of facilitating the development of the algorithm. Case studies and a performance evaluation show that the solution is suitable for fault localization in communication networks in terms of speed and reliability.


2017 ◽  
Vol 2017 ◽  
pp. 1-11
Author(s):  
Zhen Liu ◽  
Huanyu Meng ◽  
Shuang Ren ◽  
Feng Liu

Lots of multilayer information, such as the spatio-temporal privacy check-in data, is accumulated in the location-based social network (LBSN). When using the collaborative filtering algorithm for LBSN location recommendation, one of the core issues is how to improve recommendation performance by combining the traditional algorithm with the multilayer information. The existing approaches of collaborative filtering use only the sparse user-item rating matrix. It entails high computational complexity and inaccurate results. A novel collaborative filtering-based location recommendation algorithm called LGP-CF, which takes spatio-temporal privacy information into account, is proposed in this paper. By mining the users check-in behavior pattern, the dataset is segmented semantically to reduce the data size that needs to be computed. Then the clustering algorithm is used to obtain and narrow the set of similar users. User-location bipartite graph is modeled using the filtered similar user set. Then LGP-CF can quickly locate the location and trajectory of users through message propagation and aggregation over the graph. Through calculating users similarity by spatio-temporal privacy data on the graph, we can finally calculate the rating of recommendable locations. Experiments results on the physical clusters indicate that compared with the existing algorithms, the proposed LGP-CF algorithm can make recommendations more accurately.


Information ◽  
2021 ◽  
Vol 12 (6) ◽  
pp. 247
Author(s):  
Wendian Zhao ◽  
Yongjie Wang ◽  
Xinli Xiong ◽  
Yang Li

With the increase and diversification of network users, the scale of the inter-domain routing system is becoming larger and larger. Cascading failure analysis and modeling are of great significance to improve the security of inter-domain routing networks. To solve the problem that the propagation principle of cascading failure does not conform to reality, a Cascading Failure Model for inter-domain routing systems with the Recovery Feedback Mechanism (CFM-RFM) is proposed in this paper. CFM-RFM comprehensively considers the main factors that cause cascade failure. Based on two types of update message propagation mechanism and traffic redistribution, it simulates the cascading failure process. We found that under the action of the recovery feedback mechanism, the cascading failure process was accelerated, and the network did not quickly return to normal, but was rather quickly and extensively paralyzed. The average attack cost can be reduced by 38.1% when the network suffers the same damage.


2020 ◽  
Vol 21 (6) ◽  
pp. 610
Author(s):  
Xiaoliang Cheng ◽  
Chunyang Zhao ◽  
Hailong Wang ◽  
Yang Wang ◽  
Zhenlong Wang

Microwave cutting glass and ceramics based on thermal controlled fracture method has gained much attention recently for its advantages in lower energy-consumption and higher efficiency than conventional processing method. However, the irregular crack-propagation is problematic in this procedure, which hinders the industrial application of this advanced technology. In this study, the irregular crack-propagation is summarized as the unstable propagation in the initial stage, the deviated propagation in the middle stage, and the non-penetrating propagation in the end segment based on experimental work. Method for predicting the unstable propagation in the initial stage has been developed by combining analytical models with thermal-fracture simulation. Experimental results show good agreement with the prediction results, and the relative deviation between them can be <5% in cutting of some ceramics. The mechanism of deviated propagation and the non-penetrating propagation have been revealed by simulation and theoretical analysis. Since this study provides effective methods to predict unstable crack-propagation in the initial stage and understand the irregular propagation mechanism in the whole crack-propagation stage in microwave cutting ceramics, it is of great significance to the industrial application of thermal controlled fracture method for cutting ceramic materials using microwave.


Sign in / Sign up

Export Citation Format

Share Document