Search

To detect Android malware samples, a malware classification model is proposed in this paper. traditional Android malware detection and identification techniques are usually divided into static analysis and dynamic analysis. static analysis of interrelated data sets classification effect is not good, dynamic analysis effect is good but the resource consumption rate is high. In the face of this problem, we made improvements based on static analysis and proposed an artificial intelligence-based solution to achieve the purpose of malicious attack software detection by extracting feature datasets from source code and then using algorithmic models for classification processing and learning. The accuracy of the data was about 94.4% after 200 rounds of training. We used the E-validation set to validate the model, and the accuracy of the E-validation set was 90%. The results of the proposed model have achieved our expected results in terms of classification accuracy.

In federated learning a global model is trained with training data geographically distributed over a number of clients. To reduce the communication cost over the expensive wide area network, clients complete multiple local iterations before synchronization. However, since the training data are non-iid, such infrequent synchronization would compromise the accuracy after model convergence. In order to tackle this problem, we propose Two-Dimensional Learning Rate Decay (2D-LRD) in this paper, which aims to improve the model performance by adaptively tuning the learning rate on two dimensions: round-dimension and iteration-dimension during the model training. That is, we gradually decrease the learning rate and decrease the learning rates of local iterations in a synchronization round with different speeds. Based on our experiments and analysis, we find that the sum of the inner product of round updates is a valuable signal for learning rate tuning. We perform evaluation and demonstrate that 2D-LRD can make great progress compared to the baseline scheme.

This paper studies the method of transmitting beam design and directivity adjustment of segmented circular transducer. The modal synthesis method is used to design the transmitting beam pattern of segmented circular transducer, and a method of adjusting the beam directivity by using the superposition of multiple orthogonal acoustic radiation poles is proposed. A three-dimensional finite element model of segmented circular transducer is established by using finite element method, and the transmitting beam directivity of the transducer is controlled by applying the optimized weight to each channel of the transducer. Finally, the segmented circular transducer is designed and manufactured, and the experiment is carried out in the anechoic water tank to verify the method of beam design and directivity adjustment of the segmented circular transducer proposed in this paper.

The construction of a strong and smart grid is inseparable from the advancement of the power system, and the effective application of modern communication technologies allows the traditional grid to better transform into the energy Internet. With the advent of 5G, people pay close attention to the application of network slicing, not only as an emerging technology, but also as a new business model. In this article, we consider the delay requirements of certain services in the power grid. First, we analyze the security issues in network slicing and model the 5G core network slicing supply as a mixed integer linear programming problem. On this basis, a heuristic algorithm is proposed. According to the topological properties, resource utilization and delay of the slice nodes, the importance of them is sorted using the VIKOR method. In the slice link configuration stage, the shortest path algorithm is used to obtain the slice link physical path. Considering the delay of the slice link, a strategy for selecting the physical path is proposed. Simulations show that the scheme and algorithm proposed in this paper can achieve a high slice configuration success rate while ensuring the end-to-end delay requirements of the business, and meet the 5G core network slice security requirements.

Detecting out-of-domain (OOD) intents is critical in a task-oriented dialog system. Existing methods rely heavily on extensive manually labeled OOD samples and lack robustness. In this paper, we propose an efficient adversarial attack mechanism to augment hard OOD samples and design a novel generative distance-based classifier to detect OOD samples instead of a traditional threshold-based discriminator classifier. Experiments on two public benchmark datasets show that our method can consistently outperform the baselines with a statistically significant margin.

UAVs are gradually playing an important role in many civilian or military applications. With the rise of its important status, the accompanying security issues have been paid more and more attention. In this paper, we summarize the problems of small UAV cluster communication networks, and introduce the implementation of secure communication scheme in detail. By expanding the communication protocol could achieve cluster autonomous communication. Not only ensuring the security of the protocol, but also greatly saves the consumption of computing resources, energy, and significantly improves the efficiency of communication because the UAV is Resource-Constrained.

In this paper, we propose a new bus arbitration algorithm based on game theory. Since the arbitration algorithm is essentially to deal with the conflict relationship between multiple processors, and game theory is also a professional theory of contradictions and conflicts, it is logic to take the game theory into the bus arbitration algorithm. Firstly, according to the system framework, the contention of multiple processors on the bus is abstracted into a multi-person non-cooperative game, and a game model is established accordingly. Secondly, according to this model, formulate the income function to find the equilibrium point of the game. Finally, through experiments for comparison and verification, the analysis shows that the performance of the game algorithm is better than the classic arbitration algorithm.

This paper mainly discusses several problems in the estimation of the two-dimensional Direction of Arrival (DOA). The system studies the factors affecting resolution and azimuth estimation, and at the same time explores the effect of array structure on direction finding performance. In order to study the direction finding performance in the plane, the Dimension Reduction-MUSIC (DR-MUSIC) algorithm is proposed. This algorithm can effectively reduce the complexity of the traditional 2D-MUSIC algorithm and has good direction finding resolution. The validity of the algorithm is demonstrated by simulation, and a solid foundation is laid for further research on the direction of arrival estimation.

The use of artificial intelligence technology for technical and tactical analysis has become an important means and method in karate competitions and training. The integration of artificial intelligence technology and sports technical and tactical analysis is an important way to innovate and improve the technical and tactical level. In this paper, a knowledge graph in the karate domain is designed. In order to effectively use the domain knowledge graph, a new two-flow graph convolutional network is adopted, which includes classifier branch and attribute feature branch. The attribute feature branch maps the attribute embedding and score of each video instance to an attribute feature space. The semantic word embedding vectors of all concepts are used as the input of the classifier branch, and finally a classifier for behavior classification is generated. Finally, the generated classifier is evaluated according to the attribute characteristics of each video, and the whole network is optimized by the classification loss pair. In addition, the time information of the video is modeled by the attention mechanism. This technology effectively solves the problems of high labor cost, serious data loss, long delay, and low accuracy in traditional methods. The results show that the new model has a significant effect on improving the accuracy of behavior recognition, and it also lays a foundation for intelligent technical and tactical analysis. basis.

Due to the change of flight altitude and attitude of UAV, the object scale in UAV images exists difference which leads to a great challenge for object detection and has drawn wide attention. In this paper, an improved object detection network named ECascade-RCNN is proposed to deal with the multi-scale problem in object detection task for UAV images. We present an innovative Trident-FPN backbone to extract features and design a new attention mechanism to enhance the performance of the detector. Moreover, k-means algorithm is adapted to generate anchors so that the detection model can get better regression accuracy. We evaluate the proposed ECascade-R-CNN on Visdrone dataset through several ablation experiments and the results show that the ECascade-RCNN given in the paper is effective. The ECascade-RCNN is also used in the Visdrone2020 challenge and ranked 8th on the object detection track.

The translation work of business English is becoming more and more complicated, and the accuracy of translation cannot be maintained at a high level. Therefore, it is necessary to use an online proofreading system to make up for the lack of English translation software. However, due to the traditional online proofreading system, the design of the translation logic is relatively fragmented, resulting in the constant changes in the proofreading node layout and affecting the proofreading results. Therefore, an online proofreading system for business English translation based on machine vision is designed. In terms of hardware, connect a new type of single-chip microcomputer in the system and redesign the control circuit. On the software side, it calculates vocabulary similarity by tracking business English vocabulary; calculating semantic similarity based on content characteristics, and designing the system's online proofreading logic based on machine vision theory, and then resetting the system's online intelligent proofreading method. Experimental results: Compared with the system under the traditional design, the proofreading node of the design system deployment and control has a certain regularity. After proofreading the English translation, the accuracy of the translated content has been greatly improved.

A network embedding consists of a vector representation for each node in the network. Its usefulness has been shown in many real-world application domains, such as social networks and web networks. Directed networks with text associated with each node, such as software package dependency networks, are commonplace. However, to the best of our knowledge, their embeddings have hitherto not been specifically studied. In this paper, we propose PCTADW-1 and PCTADW-2, two algorithms based on neural networks that learn embeddings of directed networks with text associated with each node. We create two new node-labeled such networks: The package dependency networks in two popular GNU/Linux distributions, Debian and Fedora. We experimentally demonstrate that the embeddings produced by our algorithms resulted in node classification with better quality than those of various baselines on these two networks. We observe that there exist systematic presence of analogies (similar to those in word embeddings) in the network embeddings of software package dependency networks. To the best of our knowledge, this is the first time that such systematic presence of analogies is observed in network and document embeddings. We further demonstrate that these network embeddings can be novelly used for better understanding software attributes, such as the development process and user interface of software, etc.

Lithography enabled nanoscale fabrication in the semiconductor industry, is also the primarily driven force for the microelectronics revolution. In the history of photolithography, scientists have been working on reducing the light wavelength for higher lithographic resolution. With the rapid development of semiconductors, today's 193-nm immersion lithography is approaching its physical limits. Extreme ultra-violet (EUV) lithography with a wavelength of 13.5 nm has been emerged as the most promising candidate to continue Moore's law. However, the dramatic decrease of the wavelength of the light source has brought many technological challenges on both the machine side and the photoresist side for semiconductor high volume manufacturing (HVM). Historically, the invention of chemically amplified resists (CARs) with significantly improved sensitivity had helped overcome the power output challenge of the light source in deep ultra-violet (DUV) lithography. A possible solution for current EUVL challenges is continually developing new photoresists to efficiently absorb the EUV light, even though polymer-based CARs have been widely used for decades. This review discusses the challenges in EUVL and the progress in the EUV photoresists materials, including polymer, molecular glass, and metal-oxide nanoparticles type photoresists. A brief discussion of the development strategies and future challenges for EUV photoresists is also included.

In this paper, we have studied a novel continuously variable transmission (CVT), which has the potential to be applied to robotics. The transmission principle of the CVT is based on Vbelt transmission and planetary gear train transmission to get large reduction ratio, and its design motivation employs the theory of nested structure to realize a compact CVT. Next, we carry out dynamic simulation on the CVT with different loads and different transmission ratios to analyze work efficiency, and then analyze the dynamics of the planet carrier under the constant torque input. Finally, we develop a prototype with hydraulic drive to adjust the transmission ratio, and design a V-belt composed of multiple metal sheets that can be pre-tightened.

This paper proposes a modular two-degree-of-freedom spherical joint, which has a compact internal structure and high integration. Firstly, a six-degree-of-freedom manipulator is designed. Its core part is a two-degree-of-freedom spherical joint, and its internal structure is described in detail. Then, the mathematical modeling of the brushless DC(BLDC) motor in the joint is carried out, and the PID self-tuning control system of the BLDC motor based on the RBF neural network is designed. The control system uses the RBF neural network to identify the Jacobian information of the motor dynamic process, and adjusts the PID controller parameters in real time through Jacobian information and gradient descent algorithm, so as to optimize the performance of the entire control system. Finally, the RBF PID self-tuning control system is simulated. The simulation results show that the control system has excellent dynamic performance and steady-state performance.

Mooring location has become increasingly important with the rapid development and utilization of marine resources. The mooring system of the underwater mobile robot is required to be efficient and intelligent. This paper proposes a novel anchoring robot by imitating the excellent inhabitation functions and physiological structure of the razor clam. This bio-inspired robot can autonomously dig holes, anchor, and detach to improve the efficiency of mooring location in marine operations. Specifically, the axe foot of the robot plays an important role in these movements. It contains a screw-cone drill and two soft parts. The screw-cone drill is the main forced part while the robot is in the digging state and the rationality of drill design is proved by mechanics analysis. Moreover, the soft part, which is connected with the upper end of the drill, is analyzed to show the capability of obstacle avoidance. The rationality of the structural design and a good submarine anchoring function of the axe foot are verified by the above analysis.