Federated learning is a method of distributed machine learning based on privacy protection, which is characterized by a large number of heterogeneous clients, and often these clients have great differences in communication and computation capabilities, such as servers and smartphones. In this work, we propose a new federated learning algorithm named AAFL, which is based on asynchronous communication and can adaptively assign training models based on the capabilities of different clients to solve the communication efficiency and privacy problems caused by a large number of heterogeneous clients. Our proposed solution can assign training models with different computational complexity according to clients with different capabilities, and reduce the problem that some clients with poor capabilities slow down the training speed of the global model by means of synchronous communication. AAFL can allocate sub-models of different sizes and still aggregate a single global model. The experimental results show that assigning models of different scales according to the capabilities of different clients and using asynchronous communication can improve the training speed and accuracy of the global model to a certain extent, and has a good effect on protecting the privacy of user data.
Federated Learning (FL) is a special distributed machine learning environment. It is jointly trained by many clients under the coordination of a central server. And differential privacy can provide privacy guarantee for FL. While, federated learning, compared with centralized learning, converges at slower speed. And differential privacy exacerbates this trend. In this paper, we propose a novel FL algorithm named DP-FedADMM to solve these problems. We merge differential privacy (DP) into the FedADMM algorithm and propose a method to handle noisy gradients. Under the guidance of the help of the latest results in differential privacy theory, we provide a privacy proof of DP-FedADMM. Through extensive experiments on datasets, it is demonstrated that DP-FedADMM outperforms the currently popular DP-FedAvg algorithm in terms of model convergence speed.
The phenomenon of traffic congestion is now widespread in the traffic of major and medium-sized cities. This study creates a traffic flow diversion system with a backend command, control system management, and a data monitoring visualization large screen surveillance side to enhance operation management, incident response effectiveness, and traffic system. The system leverages relevant expertise in machine learning and visual analytics for target tracking, violation detection, and other tasks. Jetson NX receives deep learning models from Deepstream [2] to enable secure two-way communication between the edge and the cloud. The system also uses signal lights to implement intelligent traffic guidance. In this study, we illustrate the simulation using Pygame, and the simulation outcomes show the efficacy of scheduling.
Visually impaired person cannot perform their activities like ordinary person. Therefore, it becomes more difficult when the visually impaired people working with the computers in their daily life and study. In order to overcome this difficulty, an electronic braille system based on weighted Viterbi algorithm is proposed in this paper. The proposed electronic Braille system including braille input module, speech translation module and text translation module. Along with this system, the electronic braille input, two-way conversion between Braille and Chinese characters, and speech conversion can be realized. The hidden Markov model (HMM) is used to design and implement the transformation, and a weighted Viterbi algorithm is used to solve the model design details and methods. The results presented in this paper show a proper weight on the information provided by static parameters can substantially reduce the error rate, and improve color space conversion accuracy. Finally, it is shown that the system developed in this paper can help visually impaired people for their input without speech assistance, and it can help the visually impaired to study and work better.
As a defence to the SAT-based de-camouflaging attack, a brand new camouflaging strategy (called CamoPerturb) has been proposed. This camouflaging strategy achieves good resistance against the SAT-based attack. However, there is no stable generation algorithm for the perturbation circuit of the CamoPerturb. Furthermore, there is no way to control three key logic: specific output, specific gate and specific minterm, which makes it difficult for the technology to be used in practical circuits. In this paper, a perturbation circuit generation algorithm is proposed. By dividing a limited region, introducing the original input signal and narrowing the minterm range affected by the logic gate, we can find a single minterm perturbation quickly. Furthermore, we can design key logic arbitrarily. The experimental results in ISCAS’ 89 benchmark circuits and the controllers of the OpenSPARC microprocessor show that our method can be applied to the actual circuit stably with very little time consumption. Finally, our method is experimentally analyzed on an IP (intellectual property) core circuit (full adder circuit).
The increasing number of software vulnerabilities pose serious security attacks and lead to system compromise, information leakage or denial of service. It is a challenge to further improve the vulnerability detection technique. Nowadays most applications are implemented using C/C++. In this paper we focus on the detection of overflow vulnerabilities in C/C++ source code. A novel scheme named VulMiningBGS (Vulnerability Mining Based on Graph Similarity) is proposed. We convert the source code into Top N-Weighted Range Sum Feature Graph (TN-WRSFG), and graph similarity comparisons based on source code level can be effectively carried on to detect possible vulnerabilities. Three categories of vulnerabilities in the Juliet test suite are used, i.e., CWE121, CWE122 and CWE190, with four indicators for performance evaluation (precision, recall, accuracy and F1_score). Experimental results show that our scheme outperforms the traditional methods, and is effective in the overflow vulnerability detection for C/C++ source code.
In view of the complexity of electromagnetic spectrum in urban environment, detecting the UAV data transmission and image transmission signals, and the high false alarm rate, digital channelization idea is adopted. The principle is that the receiving bandwidth is divided into several channels through a low-pass filter bank, then, the polyphase components of decimation filter and the lowpass filter are used as input of polyphase filter bank. Therefore, we can achieve low-speed data flow processing. At the same time, each channel independently detects UAV data transmission and image transmission signals, with a constant false alarm algorithm setting the adaptive threshold. Experiments show that our method has achieved rapid detection of UAV signals in complex electromagnetic environments.
Highway Monitoring System is an important part of The Intelligent Highway. And the basic highway monitoring system consists of fisheye camera and gun camera as the data acquisition side. Since the gun camera has a limited viewing angle, the fisheye camera is used as a supplement. To eliminate these blind areas in the monitoring system, we propose a system that merge the fisheye image and the two guns images to obtain a global view under the monitor pole. At the same time, it provides video cutting and video annotation functions to meet the practical needs of users in use. Our experimental results show that the system provides high convenience in terms of performance and functionality.
In medical clinical applications, we must not only ensure the accuracy of classification, but also ensure that the size of the network can be carried into hardware devices with limited computing resources. In this paper, we propose an evolutionary multiobjective neural architecture search (EMNAS) framework for organ image classification. In EMNAS, evolutionary algorithms are used to encode and search network structures, which can achieve more flexible hierarchical extraction of scene information from organ med-ical images. In addition, we utilize the evolutionary multi-objective optimization to search for the Pareto optimal solutions for two conflicting objectives of computational complexity and classification accuracy. Compared with many hand-designed classical convolutional neural networks and other well-known NAS algorithms, the effectiveness of EMNAS is proved by the fact that EMNAS can achieve the best results.
Used power batteries will cause severe damage to the environment if recycling them in an unstandardized manner, and it is necessary to study how to build an effective recycling system for used power batteries. A mixed-integer programming model was constructed to minimize the cost and carbon emissions. Then, the optimal solutions obtained by genetic algorithms through examples, and studied the influence of remanufacturing rate on profit.
Environmental perception for self-driving has become a hot topic in the domain of computer vision. The field has made great progress in the past decade, and many deep learning frameworks based on single-modal data for self-driving perception tasks have emerged. However, single-modal information-based perception systems still have limitations. Recently, researchers have improved this problem using point clouds and images cross-modal fusion and obtained more satisfactory results. Therefore, it is necessary to develop a comprehensive review of recent research work. In this paper, we summarize different fusion methods, cross-modal datasets, and evaluation metrics oriented to three phases for self-driving perception tasks. Then, the point cloud and image fusion challenges are analyzed and prospected. According to the above observations, we hope that this paper can provide research ideas for self-driving point cloud and image cross-modal fusion.
The large-scale arbitrarily divisible load scheduling has been solved effectively by periodic multi-installment scheduling (PMIS), but there are relatively few studies on coarse-grained divisible loads. Therefore, we propose a coarse-grained multi-installment scheduling model in this paper, named CMIS. Two crucial issues were solved: First, an optimal load partition was derived through a strict mathematical derivation; Second, an optimal number of installments was obtained between the subtle upper and lower bounds we found. Finally, we compare the proposed model with an up-to-date model and the experimental results show that the proposed model outperforms the existing model in obtaining a shorter makespan.
With the rapid development of the social economy, traffic plays an increasingly significant role in social activities. The level of urban intelligent transportation construction directly affects the effectiveness of traffic management. This system is based on an object detection network to deal with the video monitoring, return to the target vehicle violation situation, join the Deep Sort, and be innovative to set the target tracking algorithm effectively lost judgment. The system mainly includes the following several parts: an object detection network, a target tracking algorithm, a cloud server, and a Web site. The experimental results show that the system is effective. The wide application of the system will provide powerful technical support for determining illegal traffic conditions and realizing intelligent traffic management.
News sentiment analysis is widely used in stock price forecasting, and the existing research is mostly limited to sentiment mining of news headlines, ignoring the effective information contained in article news. This study introduces extractive text summarization technology into stock price prediction, use Word2Vec and TextRank algorithm to extract effective text information contained in article news, and adopt the emotion comprehensive calculation method based on news headlines and news abstracts, taking into account the effective information of headlines and original texts. Input the comprehensively calculated news sentiment value as sentiment feature into the stock price prediction model LSTM, and propose a stock price prediction framework based on TextRank text summarization techniques and sentiment analysis. Finally, select the stock transaction data of A-share CTG DUTY-FREE for a total of 587 trading days from December 25, 2019 to May 31, 2022 for comparative experiments. The result shows, the sentiment analysis algorithm based on TextRank text summarization technology proposed in this article has the best extraction effect on the sentiment value of news texts, in terms of prediction accuracy, the model is 11.67% higher than the benchmark model on the test set.
The utilization of blockchain technology, which features a one-of-a-kind distributed architecture model, has substantially raised the bar for the reliability and traceability of information. Additionally, this technology shares many conceptual and practical parallels with the financial sharing system This topic utilizes blockchain as the underlying data technology platform, adhering to the prerequisites of decentralization, intelligence, intensification, and adaptability. The paper proposes a five-level architecture system for a financial sharing center, comprised of business activities, data collection, consensus contract, decentralized network architecture, and a sharing center. The framework system discusses the implementation and optimization of key technologies such as dynamic verification of consensus mechanisms, intelligent integration of standardized rules, and efficiency expansion.
Completeness checking of BrIM(Bridge Information Model) plays a critical role in identifying missing information according to standard documents and ensuring the depth of the designs. However, the method of traditional checking relies heavily on human work. To achieve automated completeness checking for BrIM, we propose a knowledge model construction method to solve the problem of standard specifications interpretation. Firstly, we expand the domain of the knowledge ontology based on Semantic Web technology, extract and integrate the collection of domain terms, define classes and property relationships in the ontology, and build the domain knowledge ontology. And then, we preprocess provisions to obtain single semantic rule sentences following the text characteristics of the Chinese delivery standard and construct multi-level logical expressions according to their different knowledge types. Finally, we build the domain knowledge model consisting of an ontology and logical expressions that can be used in the completeness checking of $\mathbf{BrIM}$ to describe the knowledge of codes and standards in the bridge design domain at a fine-grained level. The knowledge model improves the effectiveness and efficiency of the translation process and offers technical support for intelligent completeness checking.
At present, business education has gradually shifted from training specialized talents oriented by a certain vocational ability to training interdisciplinary and compound business talents based on industry orientation. So, vocational ability needs to be endowed with new connotation. At the macro level, the new business department provides students with interdisciplinary education with new ideas, new models, new methods and new platforms. At the micro level, the key to the success of the new business reform is to reconstruct the education and teaching system guided by the new vocational ability. This paper combs and dissolves the historical evolution of vocational ability training of business, and precisely locates the distribution point of vocational ability of new business. On this basis, the paper proposes a new construction of the vocational ability cultivation system of structured new business.
With the progress of modern science and technology, people have made great breakthroughs in the research of electronic intelligence technology. The functions of electronic products on the market can be developed according to actual needs. Because the embedded processor can be used together with a variety of sensors, it has the characteristics of low power consumption, low cost, and a stable system. It is widely used in mechanical industry production, intelligent wearable devices, and intelligent robots. The relevant embedded robot vehicles can replace human beings in post-disaster rescue and operation in a polluted environment, thus reducing the risk of personal injury and greatly improving the working conditions of people in a dangerous working environment. This paper presents a design method for Bluetooth Car Based on stm32f103c8t6 embedded processor. The design uses the main control MCU to drive the motor, and the mobile phone sends control signals to the MCU through Bluetooth to control the rotation of the steering gear, the movement direction, and the speed of the trolley. The onboard OLED displays the movement of the car. The esp32-cam camera module on the steering gear collects the image and transmits the video stream to the L/0 through Wili media.
Direction Finding of UAV has become a crucial problem at present. Traditional radio direction finding technology includes amplitude comparison direction finding and interferometer direction finding. However, when encountering coherent signals from UAVs of the same model and manufacturer, these methods cannot obtain the direction of arrival correctly or distinguish between friend and foe. Several coherent signals will be regarded as a spatial incoming signal. In this paper, received signal of uniform circular array is converted into received signal of virtual linear array through Toeplitz transformation, and then the ROOT-MUSIC algorithm is used for direction finding. Experiments show that our method can effectively solve the UAV direction finding problem.
In mobile edge computing, computing and storage resources are close to the network of the user device, which can effectively reduce network access and computing service delay. However, network function virtualization for mobile edge computing networks is more challenging due to the strong coupling of CPU resources, storage resources, and wireless resources. In order to solve the problems of load balance and long response time of user-oriented service function chain deployment algorithm in mobile edge computing networks, this paper proposed a load balance and time delay efficient algorithm (denoted as LBDQN) for virtual network function service chain deployment problem based Deep Q-LearningFirst, a global optimization model, which minimizes time delay and maximizes the load balance, is established. Then, a new real-time algorithm for VNF service chain deployment based on DQN is proposed. In this algorithm, an adaptive ε-greedy method is designed. In addition, a particle swarm optimization is used to optimize the neural network parameters. To demonstrate the efficiency of the proposed algorithm, some simulation experiments are conducted and the experimental results show that the algorithm can effectively reduce the time delay and improve the load balance than the compared algorithms.
