Network Bottlenecks Research Articles

Discovering the Representation Bottleneck of Graph Neural Networks

Discovering the Representation Bottleneck of Graph Neural Networks

Optimizing and Analyzing Traffic Data Using Data Mining Techniques

This is the project titled Optimizing and Analyzing of Traffic Data using Data Mining techniques and under this project goals will be defined which may include enhancing network perfomance ,reducing latency and strenghtening security measures. Next, critical network componemts are identified ,such as routers, switches and firewalls which require analysis and optimization. Network traffic data is collected using monitoring tools or packet sniffers, encompassing both inbound and outbound traffic. This data is then analyzed to identify patterns trends and potential security threats. Network bottlenecks and performance issues are also identified. Based on the analysis, optimizaion techniques are implemented. This may involve reconfigurring network settings, adjusting routing protocols or upgrading hard watre components, Simultaneously, security measures are implemented to address identified vulnerabilities .Perfomance improvements are continously monitored and measured using metrics such as network latency, throughput and packet loss. The optimizations are finetuned periodically based on ongoing and new traffic patterns

LIPT: Improving Prompt Tuning with Late Inception Reparameterization

Prompt tuning is a mainstream technique for fine-tuning large language models (LLMs), offering minimal parameter adjustments by learning task-specific prompt vectors. However, it suffers from training costs due to network-wide backpropagation and weaker performance compared to methods like adapters and LoRA, likely due to the limited capacity of soft prompts to encode task-specific information. This study introduces Late Inception Prompt Tuning (LIPT), a novel approach to soft prompt learning that enhances performance and efficiency by shortening backpropagation paths and employing a multidimensional bottleneck network with greater capacity. LIPT surpasses existing prompt tuning techniques on various benchmark tasks, delivering a 1.3% gain over LPT and a 5% improvement compared to standard prompt tuning when applied to RoBERTa-large, while converging more rapidly. It achieves an average accuracy of 90% across ten benchmark datasets. Notably, in certain scenarios, LIPT’s performance approaches that of full-parameter fine-tuning methods. To evaluate parameter-efficient fine-tuning (PEFT) comprehensively, we propose an Efficiency Indicator (EI) that balances accuracy and cost. LIPT is well suited for natural language understanding tasks, like sentiment analysis and text classification, with potential extensions to larger-scale models and tasks like text generation. This framework advances the scalability and practicality of fine-tuning methods for diverse applications.

An active queue management method for multiple bottleneck networks based on distributed average tracking

The active queue management (AQM) is a key technology for the information infrastructure and the Internet. The AQM for multiple bottleneck networks is especially challenging than that for single-bottleneck networks because the stability analysis of multiple bottleneck networks is more complex. In this paper, the active queue management problem for multiple bottleneck networks is considered. A novel random early detection algorithm is designed based on distributed average tracking technology, which is used to estimate the global queue length and improve the network's performance. Sufficient conditions for global asymptotic stability of the closed-loop active queue management system are presented, and some simulation results with NS-2 are given. Compared with the RED algorithm, simulation results show that the presented algorithm has smaller oscillations at routers, which highlights its effectiveness.

Network Bottlenecks and Task Structure Control the Evolution of Interpretable Learning Rules in a Foraging Agent.

Developing reliable mechanisms for continuous local learning is a central challenge faced by biological and artificial systems. Yet, how the environmental factors and structural constraints on the learning network influence the optimal plasticity mechanisms remains obscure even for simple settings. To elucidate these dependencies, we study meta-learning via evolutionary optimization of simple reward-modulated plasticity rules in embodied agents solving a foraging task. We show that unconstrained meta-learning leads to the emergence of diverse plasticity rules. However, regularization and bottlenecks in the model help reduce this variability, resulting in interpretable rules. Our findings indicate that the meta-learning of plasticity rules is very sensitive to various parameters, with this sensitivity possibly reflected in the learning rules found in biological networks. When included in models, these dependencies can be used to discover potential objective functions and details of biological learning via comparisons with experimental observations.

YOLO-DroneMS: Multi-Scale Object Detection Network for Unmanned Aerial Vehicle (UAV) Images

In recent years, research on Unmanned Aerial Vehicles (UAVs) has developed rapidly. Compared to traditional remote-sensing images, UAV images exhibit complex backgrounds, high resolution, and large differences in object scales. Therefore, UAV object detection is an essential yet challenging task. This paper proposes a multi-scale object detection network, namely YOLO-DroneMS (You Only Look Once for Drone Multi-Scale Object), for UAV images. Targeting the pivotal connection between the backbone and neck, the Large Separable Kernel Attention (LSKA) mechanism is adopted with the Spatial Pyramid Pooling Factor (SPPF), where weighted processing of multi-scale feature maps is performed to focus more on features. And Attentional Scale Sequence Fusion DySample (ASF-DySample) is introduced to perform attention scale sequence fusion and dynamic upsampling to conserve resources. Then, the faster cross-stage partial network bottleneck with two convolutions (named C2f) in the backbone is optimized using the Inverted Residual Mobile Block and Dilated Reparam Block (iRMB-DRB), which balances the advantages of dynamic global modeling and static local information fusion. This optimization effectively increases the model’s receptive field, enhancing its capability for downstream tasks. By replacing the original CIoU with WIoUv3, the model prioritizes anchoring boxes of superior quality, dynamically adjusting weights to enhance detection performance for small objects. Experimental findings on the VisDrone2019 dataset demonstrate that at an Intersection over Union (IoU) of 0.5, YOLO-DroneMS achieves a 3.6% increase in mAP@50 compared to the YOLOv8n model. Moreover, YOLO-DroneMS exhibits improved detection speed, increasing the number of frames per second (FPS) from 78.7 to 83.3. The enhanced model supports diverse target scales and achieves high recognition rates, making it well-suited for drone-based object detection tasks, particularly in scenarios involving multiple object clusters.

Panax Notoginseng Saponins promotes the meningeal lymphatic system-mediated hematoma absorption in intracerebral hemorrhage

BackgroundHematoma clearance is crucial for treating intracerebral hemorrhage (ICH). Currently, there is a lack of pharmacological therapy aimed at promoting hematoma absorption. Meningeal lymphatic system, as a drain of brain, is a potential therapeutic approach in ICH. Panax Notoginseng Saponins (PNS), proven to promote lymphangiogenesis in periphery, effectively reduces hematoma in ICH patients. However, the potential pharmacological effect of PNS on meningeal lymphatic vessels (MLVs) remains unknown. PurposeIn this study, we aimed to investigate the impact of PNS on the meningeal lymphatic system and ICH. MethodsThe collagenase-ICH model was conducted to investigate the effect of PNS. Behavioral tests, including modified neurological severity score (mNSS) and foot-fault test, and hematoma volume were used to estimate the neurological function and curative effect. The structure and drainage function of MLVs was detected by immunohistochemical staining. Visudyne intracisternal magna injection combined with red laser photoconversion was performed to ablate MLVs. RNA-sequencing was used to obtain mRNA profiles for mechanistic investigation. ResultsThe meningeal lymphatic drainage function was enhanced after ICH on day 14 without obvious lymphangiogenesis. Additionally, PNS further facilitated the process of drain with simultaneously inducing lymphangiogenesis. Moreover, ablation of MLVs by photoconverting of visudyne significantly blocked the benefits of neurological deficits improvement and hematoma absorption conducted by PNS. Furthermore, RNA-sequencing revealed that PNS regulated axonogenesis and inflammation, relying on the intact MLVs. In which, solute carrier family 17 member 7 (Slc17a7) and tumor necrosis factor (Tnf) were identified as bottleneck and hub nodes of the protein-protein interaction network of target genes, respectively. ConclusionPNS might be effective for ICH treatment by enhancing lymphangiogenesis and the meningeal lymphatic drainage function, thereby attenuating inflammation and promoting neurological recovery. The role of PNS in regulation of MLVs was investigated for the first time. This study provides a novel insight for PNS in the medical therapy of ICH.

Determination of scenarios for the implementation of projects for the restoration of transport infrastructure facilities in the post-war period based on traffic flow modeling

The author analyzes the state of the art in theory and practice. It has been established that in the context of post-war restoration of transport infrastructure, it is important to ensure effective planning and implementation of transport infrastructure restoration projects for a given project environment. The purpose of the study is to substantiate an approach to determining scenarios for the implementation of transport infrastructure rehabilitation projects based on traffic flow modeling and, on its basis, to determine a rational scenario for the implementation of transport infrastructure rehabilitation projects for a given project environment. The approach to determining scenarios for the implementation of transport infrastructure rehabilitation projects is substantiated. It is based on the modeling of traffic flows using the Greenshield model and the OpenStreetMap (OSM) framework. The Greenshield model describes the relationships between speed, density, and traffic flow. It allows to identify bottlenecks in the transportation network and prioritize areas for rehabilitation. It is proposed to use the OpenStreetMap (OSM) framework, which provides access to open data on transport infrastructure, allowing the creation of detailed models of transport networks. Based on the proposed approach, a rational scenario for the implementation of transport infrastructure rehabilitation projects for a given project environment – on the priority streets of Kramatorsk, Donetsk region - has been determined. According to the modeling results, it was found that Scenario_1.1 showed the best results in terms of average speed (31.46 km/h) and average delay (3.79 sec/km) of vehicle traffic. The obtained results confirm the feasibility of using an approach to determining scenarios for the implementation of transport infrastructure rehabilitation projects based on traffic flow modeling. Further research in this area should be conducted in the direction of developing a decision support system based on the developed approach.

Data-driven bottleneck detection on Tehran highways

In metropolitan areas, traffic congestion has become a prevalent challenge due to rapid urbanization and increased vehicle usage, adversely impacting mobility, productivity, and quality of life. Identifying and mitigating persistent traffic bottlenecks is crucial for developing efficient transportation systems and guiding infrastructure planning decisions. This research proposes an innovative data-driven methodology to pinpoint recurrent traffic bottlenecks in Tehran's extensive highway network, addressing the limitations of traditional traffic monitoring methods. Through data mining and image processing techniques applied to 16 months of traffic flow maps from Google Maps, diverse information is extracted, including traffic nodes, congestion hotspots, and locations with the longest queue lengths. The image processing approach involves color-based segmentation, pixel-level analysis, and machine learning algorithms to determine congestion levels across the highway network. The identified bottlenecks are validated against ground truth data from CCTV cameras, demonstrating a remarkable 92 % correlation for key identified points. The proposed approach leverages the power of advanced analytics to comprehensively analyze all major highways, including areas lacking CCTV infrastructure. The robust validation process reinforces the reliability of this data-driven solution in capturing real-world traffic dynamics. As urban mobility challenges escalate globally, the integration of diverse data sources and cutting-edge techniques will be instrumental in guiding intelligent transportation planning and policy decisions.

A convergent cross‐mapping approach for unveiling congestion spatial causality in urban traffic networks

AbstractSpatial causality in urban traffic networks explores how events or conditions in one location affect those in another. Unveiling congestion spatial causality is crucial for identifying congestion‐inducing bottlenecks in traffic networks and offering valuable insights for traffic network management and control. This study introduces the traffic‐convergent‐cross‐mapping (T‐CCM) method, a state‐space‐reconstruction approach from the dynamic system perspective, to identify causality among roads within urban traffic networks using time series data. Simultaneously, it effectively addresses the intricate challenges of uncertainty and interdependency among sensors caused by traffic flow dynamics. Empirical findings from real‐world (PeMS‐Bay area) traffic speed data validate the effectiveness of the T‐CCM method in detecting causality. This study reveals bidirectional causal effects between downstream and upstream roads in short‐term congestion generation and dissipation periods, which can pinpoint congestion origins and inform quick traffic management response. Furthermore, it elucidates the long‐term causality impacts between distant roads, particularly with regard to traveler choices and road land use attributes, guiding infrastructure investment and public transit improvements.

Machine Learning in Cybersecurity: Advanced Detection and Classification Techniques for Network Traffic Environments

In the digital age, the integrity of business operations and the smoothness of their execution heavily depend on cybersecurity and network efficiency. The need for robust solutions to prevent cyber threats and enhance network functionality has never been more critical. This research aims to utilize machine learning (ML) techniques for the meticulous analysis of network traffic, with the dual goals of detecting anomalies and categorizing network activities to bolster security and performance. Employing a detailed methodology, this study begins with data preparation and progresses through to the deployment of advanced ML models, including logistic regression, decision trees, and ensemble learning techniques. This approach ensures the accuracy of the analysis and facilitates a nuanced understanding of network dynamics. Our findings indicate a notable enhancement in identifying network inefficiencies and in the more accurate classification of network traffic. The application of ML models significantly reduces network delays and bottlenecks by providing a strong defence strategy against cyber threats and network shortcomings, thereby improving user satisfaction, and boosting the organizational reputation as a secure and effective service layer. Conclusively, the research highlights the pivotal role of machine learning in network traffic analysis, offering innovative insights and fresh perspectives on anomaly detection and the identification of malicious activities. It lays a foundation for future explorations and acts as an evaluation benchmark in the fields of cybersecurity and network management.

Compression-based Aggregation and Clustering Techniques for XML Enhanced Performance

This paper has presented a comprehensive literature study and discusses the primary problem of network congestion and bottlenecks in XML applications. The explosive growth of Web services has caused a significant problem in network traffic due to the largely repetitive XML textual structures. The challenge of XML message size reduction in web services has been examined, focusing on new approaches such as compression methods and message clustering for Aggregation and compression of XML messages efficiently. The goal is to examine how different strategies affect network traffic. This overview discusses twenty-five publications categorized into four: compression techniques, aggregation approaches, clustering for web services aggregation approaches, and other web services clustering models. In comparison to all models, evaluation has shown that web service clustering solutions efficiently reduce traffic. Technically, the web service clustering for aggregation models has achieved an excellent average compression ratio of up to (14.12).

SIMCB-Yolo: An Efficient Multi-Scale Network for Detecting Forest Fire Smoke

Forest fire monitoring plays a crucial role in preventing and mitigating forest disasters. Early detection of forest fire smoke is essential for a timely response to forest fire emergencies. The key to effective forest fire monitoring lies in accounting for the various levels of forest fire smoke targets in the monitoring images, enhancing the model’s anti-interference capabilities against mountain clouds and fog, and reducing false positives and missed detections. In this paper, we propose an improved multi-level forest fire smoke detection model based on You Only Look Once v5s (Yolov5s) called SIMCB-Yolo. This model aims to achieve high-precision detection of forest fire smoke at various levels. First, to address the issue of low precision in detecting small target smoke, a Swin transformer small target monitoring head is added to the neck of Yolov5s, enhancing the precision of small target smoke detection. Then, to address the issue of missed detections due to the decline in conventional target smoke detection accuracy after improving small target smoke detection accuracy, we introduced a cross stage partial network bottleneck with three convolutional layers (C3) and a channel block sequence (CBS) into the trunk. These additions help extract more surface features and enhance the detection accuracy of conventional target smoke. Finally, the SimAM attention mechanism is introduced to address the issue of complex background interference in forest fire smoke detection, further reducing false positives and missed detections. Experimental results demonstrate that, compared to the Yolov5s model, the SIMCB-Yolo model achieves an average recognition accuracy (mAP50) of 85.6%, an increase of 4.5%. Additionally, the mAP50-95 is 63.6%, an improvement of 6.9%, indicating good detection accuracy. The performance of the SIMCB-Yolo model on the self-built forest fire smoke dataset is also significantly better than that of current mainstream models, demonstrating high practical value.

Flexible Quantum Network Coding by Using Quantum Multiplexing

AbstractQuantum network coding (QNC) aims at alleviating quantum communication congestion in quantum networks. Although several QNC protocols have been presented, they cannot meet the practical requirements that part of source nodes intend to transmit their quantum states with same or different qubit numbers via the bottleneck network simultaneously. Here, the study presents a flexible QNC protocol by using quantum multiplexing. First, the entangled pairs are generated between adjacent nodes in a heralded way by using quantum multiplexing. Then the quantum memories of the source nodes and the ones of the corresponding target nodes are entangled when the intermediate nodes execute multiple rounds of entanglement swapping operations on their quantum memories. Finally, the quantum states are transmitted from the source nodes to their corresponding target nodes by means of quantum teleportation. Compared with the existing protocols, the protocol allows an arbitrary part of the source nodes to transmit their quantum states with same or different qubit numbers via the bottleneck network simultaneously, thereby exhibiting its flexibility.

Exploration On the Construction of Smart Campus All-Optical Network

With the rapid development of information technology, the construction of education informatization is urgent, and it is particularly important to promote the gradual development of colleges and universities towards digital transformation. Among them, the construction of smart campus has become an important means to enhance the level of education informatization and improve the quality of education and teaching. As an important part of the new generation of information technology, all-optical network has been widely used in the construction of smart campus. This paper first summarizes the existing network bottlenecks of traditional networks under the trend of digital campus transformation, including the problems of operation and maintenance management, network bandwidth, service expansion and weak current security, then introduces the necessity of all-optical network construction from the aspects of network bandwidth, stability and maintenance cost, and finally demonstrates the construction value of all-optical network through application scenarios. It provides a useful reference for the construction of smart campus all-optical network.

A New Generation of Collaborative Immersive Analytics on the Web: Open-Source Services to Capture, Process and Inspect Users’ Sessions in 3D Environments

Recording large amounts of users’ sessions performed through 3D applications may provide crucial insights into interaction patterns. Such data can be captured from interactive experiences in public exhibits, remote motion tracking equipment, immersive XR devices, lab installations or online web applications. Immersive analytics (IA) deals with the benefits and challenges of using immersive environments for data analysis and related design solutions to improve the quality and efficiency of the analysis process. Today, web technologies allow us to craft complex applications accessible through common browsers, and APIs like WebXR allow us to interact with and explore virtual 3D environments using immersive devices. These technologies can be used to access rich, immersive spaces but present new challenges related to performance, network bottlenecks and interface design. WebXR IA tools are still quite new in the literature: they present several challenges and leave quite unexplored the possibility of synchronous collaborative inspection. The opportunity to share the virtual space with remote analysts in fact improves sense-making tasks and offers new ways to discuss interaction patterns together, while inspecting captured records or data aggregates. Furthermore, with proper collaborative approaches, analysts are able to share machine learning (ML) pipelines and constructively discuss the outcomes and insights through tailored data visualization, directly inside immersive 3D spaces, using a web browser. Under the H2IOSC project, we present the first results of an open-source pipeline involving tools and services aimed at capturing, processing and inspecting interactive sessions collaboratively in WebXR with other analysts. The modular pipeline can be easily deployed in research infrastructures (RIs), remote dedicated hubs or local scenarios. The developed WebXR immersive analytics tool specifically offers advanced features for volumetric data inspection, query, annotation and discovery, alongside spatial interfaces. We assess the pipeline through users’ sessions captured during two remote public exhibits, by a WebXR application presenting generative AI content to visitors. We deployed the pipeline to assess the different services and to better understand how people interact with generative AI environments. The obtained results can be easily adopted for a multitude of case studies, interactive applications, remote equipment or online applications, to support or accelerate the detection of interaction patterns among remote analysts collaborating in the same 3D space.

Application of Graph Theory for Blockchain Technologies

Blockchain technology, serving as the backbone for decentralized systems, facilitates secure and transparent transactional data storage across a distributed network of nodes. Blockchain platforms rely on distributed ledgers to enable secure peer-to-peer transactions without central oversight. As these systems grow in complexity, analyzing their topological structure and vulnerabilities requires robust mathematical frameworks. This paper explores applications of graph theory for modeling blockchain networks to evaluate decentralization, security, privacy, scalability and NFT Mapping. We use graph metrics like degree distribution and betweenness centrality to quantify node connectivity, identify network bottlenecks, trace asset flows and detect communities. Attack vectors are assessed by simulating adversarial scenarios within graph models of blockchain systems. Overall, translating blockchain ecosystems into graph representations allows comprehensive analytical insights to guide the development of efficient, resilient decentralized infrastructures.

Cooperative Distributed Uplink Cache over B5G small cell networks.

The emergence of content-centric network has resulted in a substantial increase in data transmission in both uplink and downlink directions. To tackle the ensuing challenges of network congestion and bottlenecks in backhaul links within Beyond Fifth Generation (B5G) networks, data caching has emerged as a popular solution. However, caching for uplink transmission in a distributed B5G scenario poses several challenges, including duplicate content matching and users' obliviousness about cached contents. Furthermore, it is important to maximize available space by caching the most popular contents in a distributed manner. In this paper, we propose two schemes for uplink transmission in distributed B5G SCNs. The first scheme focuses on content matching to eliminate duplicate contents among distributed caches, while the second scheme redistributes un-duplicated cached contents among distributed caches based on their available space and content's size. These approaches aim to enhance energy and spectral efficiency by reducing unnecessary uploads and optimizing distributed content caching, in addition to improve the content delivery. The analysis shows that the proposed schemes outperform the existing schemes by improving the cache hit ratio, cache hit probability, overall distributed cache efficiency, and diversity by 29.17%, 74.89%, 24.17%, and, 80%, respectively. Furthermore, the average throughput, Spectrum Efficiency (SE), and Energy Efficiency (EE) of the access network is improved by 17.78%, 18%, and 78%, respectively. Besides that, the EE and SE of both the sidehaul and backhaul links of the SBSs are also improved.

Expanding bottlenecks reveals hidden bottlenecks and leads to more congested city centers

Urban traffic congestion is a perennial and critical issue. During the morning and evening rush hours, a large number of people flood into the urban road network within a short period of time, resulting in severe traffic congestion and significant inconvenience to daily life. Bottlenecks are weak links of the road network and congestion frequently occurs at these points. Many researchers postulate that unblocking bottlenecks can enhance the traffic environment. However, this perspective overlooks the propagation characteristics of congestion. To mitigate traffic congestion, our study focuses on the traffic flow propagation process post-expansion of bottlenecks in urban road networks. Our findings indicate that merely unblocking bottlenecks may lead to an increased influx of suburban residents into city centers for work within a short period of time. It leads to a surge in vehicle density in the central area, which can cause severe traffic congestion problems and make significant increase in the length of commuting time. Additionally, we identify numerous hidden bottlenecks within the road network that become active and transform into new bottlenecks once existing bottlenecks are unblocked. We simulate the commuting behavior of residents during morning rush hours on a simplified road network to quantify the adverse effects of bottleneck expansion. A large-scale simulation of the Beijing road network supports our finding that simply unblocking bottlenecks just only greatly increases residents’ commuting time. This research can provide an in-depth understanding of traffic congestion problems’ mechanisms. It is very beneficial for optimizing the construction of cities.

Exploiting Label Skews in Federated Learning with Model Concatenation

Federated Learning (FL) has emerged as a promising solution to perform deep learning on different data owners without exchanging raw data. However, non-IID data has been a key challenge in FL, which could significantly degrade the accuracy of the final model. Among different non-IID types, label skews have been challenging and common in image classification and other tasks. Instead of averaging the local models in most previous studies, we propose FedConcat, a simple and effective approach that concatenates these local models as the base of the global model to effectively aggregate the local knowledge. To reduce the size of the global model, we adopt the clustering technique to group the clients by their label distributions and collaboratively train a model inside each cluster. We theoretically analyze the advantage of concatenation over averaging by analyzing the information bottleneck of deep neural networks. Experimental results demonstrate that FedConcat achieves significantly higher accuracy than previous state-of-the-art FL methods in various heterogeneous label skew distribution settings and meanwhile has lower communication costs. Our code is publicly available at https://github.com/sjtudyq/FedConcat.