Breast cancer is a major health concern affecting women worldwide. Early detection and accurate diagnosis of breast cancer are crucial for improving patient outcomes. In recent years, deep learning techniques have been increasingly applied to medical imaging, including mammography, for the detection and diagnosis of breast cancer. In this study, we proposed a deep learning-based approach using the EfficientNet architecture for the detection and classification of breast cancer. We evaluated the performance of our proposed approach using mammography images from the CBIS-DDSM dataset and achieved accuracy of 0.75 and AUC of 0.83. Our results demonstrate the effectiveness of using deep learning techniques in medical imaging for breast cancer detection and diagnosis.
Cyberspace mimic defense (CMD) is an emerging defense technology that combines dynamic heterogeneous redundancy (DHR) to enhance the security of information systems. At present, the applications of CMD focus more on data tamper resistance and do not more pay attention to the privacy disclosure of executors. This paper proposes to introduce the secret sharing mechanism into the CMD and use a secret sharing scheme with dimension switching to achieve the security of the private data of executors without adding extra key management. And experiments proved that the combination of the secret sharing scheme and mimic defense can increase data security at a tolerable cost.
Most existing methods for few-shot image classification treat the support features and query features separately, and the category of a query image is determined based on its similarity to the support images. However, the relationships between the query features and support features are often ignored. In this paper, we propose a strategy that exploits the relationships between query and support features and give more weights to the highly related features. A cross-feature fusion method is further proposed to combine with metric learning to reduce overfitting. Extensive experiments on a wide range of datasets show that our method has achieved advanced results.
Bokeh rendering is a popular and effective technique used in photography to create an aesthetically pleasing effect. It is widely used to blur the background and highlight the subject in the foreground, thereby drawing the viewer's attention to the main focus of the image. In traditional digital single-lens reflex cameras (DSLRs), this effect is achieved through the use of a large aperture lens. This allows the camera to capture images with shallow depth-of-field, in which only a small area of the image is in sharp focus, while the rest of the image is blurred. However, the hardware embedded in mobile phones is typically much smaller and more limited than that found in DSLRs. Consequently, mobile phones are not able to capture natural shallow depth-of-field photos, which can be a significant limitation for mobile photography. To address this challenge, we propose a novel method for bokeh rendering based on the Vision Transformer. The proposed method employs an adaptive depth calibration network (ADCN) that acts as a confidence level to compensate for errors in monocular depth estimation. The ADCN is used to supervise the rendering process in conjunction with depth information, allowing for the generation of high-quality bokeh images at high resolutions. Extensive experiments demonstrate that the proposed method outperforms state-of-the-art (SOTA) methods, achieving about 24.7% improvements on LPIPS and obtaining higher PSNR scores.
The employment of decentralized microservices requires a lightweight service composition scheme with flexibility. However, no schemes can fulfil the requirement. In this paper, a data and control interdependent problem in traditional API form is identified as the root preventing a scheme fulfilling the requirement from being devised. A new configurable API form, Standardized Control Data Separation API (SCONDAS API), is devised to resolve the data and control interdependent problem. An API framework implementing the form of SCANDAS API is further proposed. The proposed API framework enables lightweight and flexible microservice composition. Experimental evaluations validate the feasibility of the API framework.
The energy conservation law (ECL) is one of the most fundamental physical laws in the nature. The ECL has some different manifestation forms. The various manifestation forms can be mathematically expressed in terms of the corresponding energy theorems. The various energy theorems govern the energy utilization mechanisms of the corresponding physical phenomena. This paper simply reviews some important electromagnetic energy theorems, such as the entire-structure-oriented work-energy theorem (ES-WET), the partial-structure-oriented work-energy theorem (PS-WET), the energy transport theorem (ETT), the Poynting theorem, the Rayleigh-Carson reciprocity theorem (RT), the Rumsey RT, the Lorentz RT, and the Parseval theorem. The ES-WET governs the energy-scattering phenomenon; the PS-WET governs the lumped-port-driven energy-transmitting phenomenon; the ETT governs the wave-port-fed energy-transmitting phenomenon; the Poynting theorem governs the energy-absorbing phenomenon; the RTs govern the energy-coupling phenomenon; the Parseval theorem guarantees the energy invariance in the time domain, frequency domain, and spectrum domain. This paper simply summarizes some valuable energy-viewpoint-based modal analysis theories, such as the ES-WET-based characteristic mode theory (CMT) for energy-scattering phenomenon, the PS-WET-based CMT for lumped-port-driven energy-transmitting phenomenon, the ETT-based decoupling mode theory (DMT) for wave-port-fed energy-transmitting phenomenon, and the Poynting-theorem-based DMT for energy-absorbing phenomenon.
Compared with previous generations of traditional network, the 5G core network uses a service-based architecture, in which the communication between network functions (NF) depends on a centralized multi-level service registry and discovery mechanism. In the 6G era, the core network will be further distributed and the participants in the network ecosystem will be more diverse. Therefore, a more flexible and efficient way is needed to achieve rapid addressing while avoiding the trust problem between different participants. As a “decentralized” information storage and management technology, Blockchain is naturally suitable for solving this problem. In this paper, we propose an effective service registry and discovery method. This method uses blockchain technology to record the information of the core network NF, then other processes can be realized through it. The application scenario and scheme of this method are introduced, as well as the feasibility and functional integrity are verified.
Aiming at the problem of mobile localization in nonline-of-sight (NLOS) environments, in this paper a new mobile location method based on scattering path identification is proposed. First of all, an improved single scattering path identification algorithm which is used to identify whether two scattering paths is from the same scatterer based on line-of-possible-mobile-device (LPMD) line is proposed, then the scattering paths from the same scatterer at different BSs are grouped and the scatterer is located by these paths and BSs, at last, the located scatterers are utilized as virtual base stations to take part in the localization of mobile. Simulation results show that the proposed method can effectively improve the localization accuracy of mobile station, which is close to the accuracy acquired in line-of-sight (LOS) environments.
This paper focuses on a hierarchical MapReduce distributed computing network, where distributed workers connect with multiple relay nodes to form several clusters, and the relay nodes communicate with each other via a central server. In the MapReduce tasks, the workers are assigned to compute output functions with the help of relays and the central server. We consider the case where the server can not store data (e.g., a switch node), and propose a file placement and coded shuffle scheme to reduce the uplink and downlink communication loads.
This paper takes 14 cities in Liaoning province as research samples to analyze the economic development strength of each city. The article is based on the data related to the Statistical Yearbook of Liaoning Province in 2020 and selects 9 indicators from the Statistical Yearbook using stata for correlation analysis, regression analysis, etc. And the economic development ability of 14 cities in Liaoning province is analyzed and compared by the principal component factor method in factor analysis. Then, the cities' scores were ranked, and the 14 cities were divided into three echelons according to their scores. Finally, corresponding suggestions are made for the economic development of the cities based on the current economic development of each region.
With the development of the sixth Generation (6G) mobile communication, big data and deep learning, new requirements for latency and computational complexity have been put forward. Therefore, in this paper, we propose a new floating-point storage called extended block floating-point (eBFP). This article first briefly introduces the storage method of the IEEE Standard for Floating-point Arithmetic (IEEE-754) floating-point numbers and its shortcomings, and then details the storage method of the eBFP operator and its specific operation logic. Finally, a simulation comparison with IEEE-754 floating-point number shows the advantage of eBFP to better trade-off accuracy and complexity.
This paper aims to efficiently save energy consumption of multiple-input-single-output (MISO) networks. To this end, an optimization problem is formulated to minimize the transmit power of the base station (BS) while ensuring the minimal required information rate of each user. Reinforcement learning (RL) is a significant solution to tackle this non-convex problems. Nevertheless, designing efficient reward functions that accurately describe the involved optimization problems for RL is a challenging problem, which may result in low learning efficiency and even inaccurate learning. To cope with such an issue, this paper proposes an inverse reinforcement learning (IRL)-based approach to minimize the transmit power of MISO networks, in which the generative adversarial imitation learning (GAIL) is employed as the IRL framework and the reward function is represented by an artificial neural network (ANN) model, whose parameters are learned automatically in terms of an expert policy based on the semidefinite relaxation (SDR)-based approach. Since SDR is able to achieve the near optimal solution, the learned reward function is more accurate than traditional artificially designed one. Simulations show that, the proposed IRL-based approach reaches similar energy consumption to the SDR-based approach, and reduces about 50% transmit power compared to traditional RL-based one.
Federated learning (FL) is a distributed learning framework, which can be widely applied into wireless networks to improve network intelligence without disclosing users' private data. However, due to the scarce wireless resources, it is inefficient for massive user devices to participate in FL training simultaneously. To address this challenge, in this work, we introduce the client clustering scheme and decompose the training process into two stages. Firstly, we divide user devices into different clusters based on Density-Based Spatial Clustering of Applications with Noise (DBSCAN) and Low Energy Adaptive Clustering Hierarchy (LEACH) algorithms. Then a scoring mechanism considering user devices' locations, velocities, link retention time, and training latency is proposed to determine the cluster head among the devices in each cluster. Finally, simulation results validate the performance of the proposed client clustering method in terms of model accuracy and training latency.
In recent years, molecular communication (MC) has gradually become a research hotspot. Human microvascular network is one of the main application scenarios for MC systems. However, there are few researches on the topological model of the microvascular network at this stage, which also becomes a major technical obstacle for the further development of MC systems. In this paper, inspired by human arterial tree, we propose a MC system model in microvascular network based on binary tree structure. This model abstracts part of the blood vessel network of human body into a binary tree. The junction of blood vessels are represented by binary tree nodes. We evaluate the proposed model in terms of bit error rate (BER) by simulations. This work can pave the way for the implementation of Internet of in-body Nano Things (IoNTs).
Understanding vulnerability trends is critical to the risk management process. The goal of this paper is to raise the awareness and inform the users about the Common Vulnerabilities and Exposures (CVE) associated with Known Exploited Vulnerabilities that might affect users of the products by gathering data from the National Vulnerability Database (NVD) and CVE details from 2017 to Present. We built a system using Python and CVE Analyzer, which allow users to search for specific CVE records based on CVE ID, vendors, products, published date and the last updated date and export rows of data based on the filtered keyword. We also analyze the overall frequency of CVE records each year regarding Common Vulnerability Scoring System (CVSS) base metric trends, the vendors, and products with the most ransomware vulnerabilities. Our findings show that the frequency of all ransomware vulnerabilities increased from 2019 to 2021. Also, the number of high CVSS score of CVE records associated with ransomware is higher than those with low CVSS scores. Apple and Microsoft are the top vendors with the most exploited ransomware vulnerabilities, and IOS and Windows 10 are the products with the highest vulnerable versions by Apple and Microsoft. Our system can help users to prevent and mitigate the impact of ransomware attacks using datasets and data analysis.
[Purpose] With the rapid development of social networks, the public opinion environment on the Internet is changing rapidly. Nowadays, social networks have become one of the important platforms for internet users to understand current political news and discuss hot news, and the generation and dissemination of emergencies are inseparable from the important channel of social networks. It is necessary to effectively channel public opinion, reduce the malicious spread of users, and resolve the crisis of public opinion when an emergency occurs. [Method] This paper collects the comment data of three emergencies at the national security level under the microblog platform, and uses the natural language processing model to process the data, and constructs the emergency event evolutionary graph and the emergency abstract event evolutionary graph. [Conclusion] The results show that when a major emergency occurs, the severity, details and background of the event attract the most attention from the public, and they are the core elements that constitute the event evolutionary graph. In emergencies involving all aspects of national security, politics and medical care, the relevant departments shall guide a positive and favorable public opinion atmosphere in society and create a harmonious public opinion environment.
Isolated sign language recognition has been an important part of breaking down communication bottlenecks for deaf-mute and others. While facing this problem, the purpose of this paper is to classify American isolated sign language video by modeling pose, hands and face keypoints representation. Specifically, this paper introduces a novel framework whose main components are the altered Dense Predictive Coding (DPC) pre-trained model and the Encoder pre-trained model. The DPC model is trained using self-supervised learning to obtain representation of pose and hands keypoints. The Encoder model is trained using supervised learning to obtain representation of face keypoints. Combining the altered DPC model with image inductive biases and the Encoder model with a self-attention mechanism, the final combined model achieves 0.81 on the test set of the ISAL dataset, outperforming the current open-source solution by a significant margin.
This paper attempts to use convolutional neural networks to classify images of patients with COVID-19. The classification model is constructed based on enormous actual X-ray images of patients with COVID-19 patients and normal people. In this paper, the structure is optimized based on the Residual Neural Network model, and the accuracy is improved from 0.91 to 0.99. In addition, NasNet, DenseNet and Vgg-16, were compared horizontally in terms of accuracy, precision, recall and F1 score. ResNet reached 99.9% in the training set of accuracy, and 96.7% in the test set of accuracy, which performed well.
In this paper, we consider the TDMA uplink system, where multiple status update devices and multiple throughput service devices upload their traffics to one access point in TDMA mode, and all devices are assumed to be provisioned with data buffer to store data before transmission. In order to meet the fresh status update requirements, the maximal allowed long term average age of information (AoI) of all status update devices is specified. On the basis, the sum rate maximization problem of all throughput service devices is presented to derive the adaptive TDMA uplink transmission scheme that meets the AoI requirements. Our analysis results show that, the proposed adaptive TDMA uplink transmission scheme can effectively coordinate different performance requirements of status update and throughput service by explicitly introducing AoI requirements in the adaptive transmission design, so as to ensure heterogeneous multiple access requirements.
Over the past decade, Satellite communication has been a crucial topic. In particular, as we move from the 5G era to the 6G era, satellites will play an increasingly critical role in providing coverage and flexibility. With the increasing complexity of satellite network environment, the resource allocation of satellite network is becoming more and more important. Accurate identification of traffic types for classification can allocate network resources more effectively. Each method has its own advantages and disadvantages. In order to minimize the impact of the disadvantages, a reasonable combination of them is a new way to accomplish this task. In this paper, we propose a traffic classification method based on deep packet inspection (DPI) and convolution neural network (CNN), and verify it with open data sets. Experimental results show the effectiveness of our proposed method.
