نشریه مهندسی برق و مهندسی کامپیوتر ایران
سال بیستم شماره 4 (پیاپی 75، زمستان 1401)

Mobile edge computing (MEC) are new issues to improve latency, capacity and available resources in Mobile cloud computing (MCC). Mobile resources, including battery and CPU, have limited capacity. So enabling computation-intensive and latency-critical applications are important issue in MEC. In this paper, we use a standard three-level system model of mobile devices, edge and cloud, and propose two offloading and scheduling algorithms. A decision-making algorithm for offloading tasks is based on the greedy Knapsack offloading algorithm (GKOA) on the mobile device side, which selects tasks with high power consumption for offloading and it saves energy consumption of the device. On the MEC side, we also present a dynamic scheduling algorithm with fuzzy-based priority task scheduling (FPTS) for prioritizing and scheduling tasks based on two criteria. Numerical results show that our proposed work compared to other methods and reduces the waiting time, latency and system overhead. Also, provides the balance of the system with the least number of resources. And the proposed system reduces battery consumption in the smart device by up to 90%. The results show that more than 92% of tasks are executed successfully in the edge environment.

Multi-label classification is one of the important issues in machine learning. The efficiency of multi-label classification algorithms decreases drastically with increasing problem dimensions. Feature selection is one of the main solutions for dimension reduction in multi-label problems. Multi-label feature selection is one of the NP solutions, and so far, a number of solutions based on collective intelligence and evolutionary algorithms have been proposed for it. Increasing the dimensions of the problem leads to an increase in the search space and consequently to a decrease in efficiency and also a decrease in the speed of convergence of these algorithms. In this paper, a hybrid collective intelligence solution based on a binary particle swarm optimization algorithm and local search strategy for multi-label feature selection is presented. To increase the speed of convergence, in the local search strategy, the features are divided into two categories based on the degree of extension and the degree of connection with the output of the problem. The first category consists of features that are very similar to the problem class and less similar to other features, and the second category is similar features and less related. Therefore, a local operator is added to the particle swarm optimization algorithm, which leads to the reduction of irrelevant features and extensions of each solution. Applying this operator leads to an increase in the convergence speed of the proposed algorithm compared to other algorithms presented in this field. The performance of the proposed method has been compared with the most well-known feature selection methods on different datasets. The results of the experiments showed that the proposed method has a good performance in terms of accuracy.

Block-centric graph processing systems have received significant attention in recent years. To produce the required partitions, most of these systems use general-purpose partitioning methods. As a result, the performance of them has been limited. To face this problem, special partitioning algorithms have been proposed by researchers. However, these methods focused on traditional partitioning measures like the number of cutting-edges and the load-balance. In return, the power of block-centric graph processing systems is due to unique characteristics that are focused on the design of them. According to basic and important characteristics of these systems, in this paper two new measures are proposed as partitioning goals. To the best of our knowledge, the proposed method is the first work that considers the diameter and size of the high-level graph as optimization factors for partitioning purposes. The evaluation of the proposed method over real graphs showed that we could significantly reduce the diameter of the high-level graph. Moreover, the number of cutting-edges of the proposed method are very close to Metis, one of most popular centralized partitioning methods. Since the number of required supersteps in block-centric graph processing systems mainly depends on the diameter of the high-level graph, the proposed method can significantly improve the performance of these systems.

Fog computing is an emerging research field for providing cloud computing services to the edges of the network. Fog nodes process data stream and user requests in real-time. In order to optimize resource efficiency and response time, increase speed and performance, tasks must be evenly distributed among the fog nodes. Therefore, in this paper, a new method is proposed to improve the load balancing in the fog computing environment. In the proposed algorithm, when a task is sent to the fog node via mobile devices, the fog node using reinforcement learning decides to process that task itself, or assign it to one of the neighbor fog nodes or cloud for processing. The evaluation shows that the proposed algorithm, with proper distribution of tasks between nodes, has less delay to tasks processing than other compared methods.

Images captured in low light conditions are unsuitable for human and machine vision due to low brightness and sharpness and high noise, and have a negative effect on their performance. Much research has been done to improve such images. The methods proposed so far to solve this problem greatly improve such images. One of these methods is the RETINEX-based method, which modifies low-light images, but because the initial structure of this method is complex and inefficient, researchers have developed other methods such as SSR, MSR, and MSRCR. To solve the problem, they have presented this approach. These methods, in turn, have problems such as abnormal images and amplification of noise. In the continuation of the work done, the field of optimization has been used, which shows better performance than the previous works. In this research, by obtaining the optimal brightness component, using nonlinear conversion and applying smoothing filter and reducing noise on the image as a post-processing step, these weaknesses are largely eliminated. By applying the proposed method, the resulting images look more natural and their information is more preserved. Subjective and objective criteria such as EI, SSIM, PSNR and IMMSE were used to evaluate the proposed method. The simulation results show the superiority of the proposed method over the competing methods.

Understanding and analyzing spatial-temporal data changes is very important in various applications, including the protection and development of natural resources. In the past studies, Markov process and comparison-based methods were mainly used to predict the changes of vegetation indices, whose accuracy still needs improvement. Although time series analysis has been used to predict some indices, the method to extract these indices from remote sensing data and model their sequences with deep learning is rarely observed. In this article, a method for predicting changes in plant indices based on deep learning is presented. The research data includes Landsat satellite images from 2000 to 2018, related to four seasons in the north and east of Shahrood city in Semnan province. The time span of the extracted images makes it possible to predict changes in vegetation cover. Vegetation indices extracted from the data set are NDVI, SAVI and RVI. After performing atmospheric corrections on the images, the desired indicators are extracted and then the data is converted into a time series. Finally, the modeling of the sequence of these data is performed by the Short-Long-Term Memory (LSTM) network. The results of the experiments show that the deep network is able to predict future values with high accuracy. The amount of the model error without additional data is 0.03 for the NDVI index, 0.02 for the SAVI index, and 0.06 for the RVI index.

The increasing capabilities of integrated circuits and the complexity of applications have led hardware design methods and tools to higher levels of abstraction and high-level synthesis is one of the key steps in increasing the level of abstraction. In recent years, extensive research has been conducted on the design of decimal arithmetic reconfigurable architectures. Since, on the one hand, the effective use of these architectures depends on the existence of appropriate algorithms and tools to implement the design on the hardware, and on the other hand, research on the development of these algorithms has been very limited, this paper will present methods for the automated synthesis of decimal arithmetic circuits on a coarse-grained reconfigurable architecture. The platform chosen to execute the proposed algorithms is the DARA coarse-grained reconfigurable architecture, which is optimized for decimal arithmetic. The algorithms proposed for resource allocation of synthesis include a heuristic method and an ILP algorithm. The results show that, as expected, for the limited architectural dimensions used, the ILP algorithm performs significantly (about 30%) better than the heuristic algorithm.

One of the evaluation methods for image steganography is preserving cover image quality and algorithm imperceptibility. Placing hidden information should be done in such a way that there is minimal change in quality between the cover image and the coded image (stego image). The quality of the stego image is mainly influenced by the replacement method and the amount of hidden information or the replacement capacity. This can be treated as an optimization problem and a quality function can be considered for optimization. The variables of this function are the mappings applied to the cover image and the hidden information and location of the information. In the proposed method, by genetic algorithm and using the two concepts of targeted search and aimless search, the appropriate location and state for placement in the least significant bits of the cover image are identified. In this method, hidden information can be extracted completely and without error. This feature is important for management systems and cloud networks that use steganography to store information. Finally, the proposed method is tested and the results are compared with other methods in this field. The proposed method, in addition to maintaining the stego image quality, which is optimized based on PSNR, has also shown good performance in examining histogram and NIQE statistical criteria.