نشریه محاسبات نرم
سال دوازدهم شماره 2 (پیاپی 24، پاییز و زمستان 1402)

A large amount of research in the field of online learning has focused on the problem of overcoming catastrophic forgetting, and few research studies have focused on classifying the data stream with appropriate accuracy and running time. On the other hand, due to the volume and type of data stream, many traditional machine learning algorithms do not have the necessary efficiency when faced with it. Thus, in this paper, a novel model using reinforcement learning and the stochastic gradient descent algorithm is presented for the classification stream data with appropriate accuracy and running time. One of the important features of reinforcement learning is that the agent can adapt its behaviour gradually to the changes that occur and gradually add to its previous knowledge. In this research, because of the use of reinforcement learning and the definition of reward, the agent has a better performance in the environment. The proposed algorithm has been tested on various data, including the dataset of human activity recognition, and compared with several incremental algorithms in terms of accuracy and running time. According to the experimental results, the proposed algorithm has the best performance in terms of both accuracy and running time compared to other incremental algorithms.

‎Let ‎G ‎be a ‎‎graph ‎and ‎d_G(v) ‎be ‎the ‎degree ‎of v ‎in ‎G‎. Then, the sum of the k-power degrees of ‎‎G ‎is ‎defined ‎as‎ ‎\Sigma_k(G) = \sum_{u \in V(G)}d_G(u)^k‎‎‎. ‎In this paper‎, ‎we obtain a relationship between Stirling number‎, ‎number of trees of subsets and the sum of the k-power degrees of the chemical graphs‎. ‎Also‎, ‎we characterize the extremal unicyclic graphs based on the sum of the k-power degrees of the graphs‎.

Nowadays, with the development of digital technology, the role of computer applications is important in art, especially in fabric and cloth design. Note that in traditional methods, it is not possible to interact with the consumer until the end of the design process, and if the final design is not approved by them, all design steps must be repeated. Typical design software systems only work well for professionals and are difficult for non-professionals to work with; therefore, there is a need to use systems that can provide the interaction between the user and the system while maintaining design speed, which is essential in this field. In this research, a fabric design assistance system based on the interactive genetic algorithm has been developed. To design the fabric, patterns and colors available in Qashqai kilims, as well as the type of pattern arrangement in kilims, were used. The designs produced by the system are displayed to the user to estimate the level of fitness. According to the user's evaluation, weaker designs are discarded, and stronger designs are improved by passing through the system again, and finally, the desired design is created. The results show that the use of the proposed system in the fabric design industry enables designers, buyers, and even fabric manufacturers to apply their taste in the fabric design process. As inferred from the users' point of view, reducing the design process time, reducing related costs, and achieving multiple designs in the shortest possible time are the advantages of this system.

Home intelligence is one of the most practical and thriving topics in today’s world, which allows people to adjust and control electronic equipment remotely, as well as program them to save energy. In recent years, many companies have introduced hardware and software systems as appliances and communication technologies for smart homes. These technologies differ from each other in various aspects such as implementation conditions, cost, scalability, security, etc., and due to their great diversity, it seems difficult to choose a suitable technology under given conditions. It is quite clear that the prerequisite for the proper design and implementation of a smart home is to know the features and limitations of these technologies in different conditions. This article reviews common communication technologies in smart homes based on the communication medium, i.e., wire, power lines, and wireless, and outlines the features and limitations of each technology, as well as the related research and challenges ahead to improve such technologies. Also, the highlighted points in the article can help the designers of smart home systems select and use the appropriate technology according to the conditions in a smart home.

In this paper, we propose an automatic detection system for COVID-19 cases based on the Internet of Things. In the proposed model, first, using Internet of Things technology, medical images are sent directly to the data collection after the suspicious person's visit through medical equipment equipped with Internet of Things, and then, in order to help radiologists to interpret medical images better, usage has been made of four pre-trained convolutional neural network models i.e. InceptionV3, InceptionResNetV2, VGG19 and ResNet152 as well as two datasets of chest radiology medical images and CT Scan in a 3-class classification for accurate prediction of cases suffering from COVID-19, healthy people, and diseased cases. Finally, the best result for CT-Scan images is related to InceptionResNetV2 architecture with an accuracy of 99.366%, and for radiology images related to the InceptionV3 architecture, it is 96.943%. The results show that this system leads to a reduction in daily visits to medical centers and thus reduces the pressure on the medical care system. It also helps rheology specialists to identify the disease as quickly as possible.

Among the topics in the field of software engineering, energy efficiency is an influential factor in software development and maintenance, especially for battery-limited devices. Although software refactoring can improve software quality, recent studies suggest that applying some refactoring operators may lead to conflicts with energy consumption and execution time of Android applications. In this paper, we analyze the impact of code refactoring for eight Android/Java bad code smells and anti-patterns. To conduct the studies and obtain the results, we use a testbed of five real and one synthetic Android applications. In the first step, we measure energy consumption, execution time and quality design of application before and after refactoring. The results show that in some cases, refactoring leads to a decrease in energy consumption and execution time, and in others, it increases energy consumption and application execution time. We then propose a novel refactoring recommendation approach based on evolutionary multi-objective optimization that accounts for energy consumption, execution time and refactoring effort for Android/Java anti-patterns.  For this purpose, we use Nondominated Sorting Genetic Algorithm-II (NSGA-II) with three objectives 1) energy consumption, 2) execution time, and 3) refactoring effort. The obtained results show that this approach can generate refactoring recommendations with a median precision of 65% and 76% for improving energy and execution time, respectively, while the median of removed antipatterns in testbed applications is 42%.

Fog computing is increasingly used as a platform for processing Internet of Things applications. Thus, this architecture extends cloud computing services to the edge of the network, where processing may be cheaper and faster. One of the main challenges in providing Quality of Service (QoS) requirements, such as delay and energy consumption in the fog environment, is to pay attention to the energy limitation and low computational capacity of fog nodes, which makes it difficult to assign tasks to fog nodes. This paper first presents a mathematical model for resource allocation with the aim of reducing delay and energy while considering QoS criteria. Then, a combined genetic and grey wolf algorithm is introduced to solve the model. Note that the combination of these two algorithms leads to finding an optimal solution efficiently. Although the implementation of the proposed algorithms has processing costs and computational delay, due to the improvement of QoS criteria, this cost can be ignored. The results show that the combination and simultaneous use of the positive points of the two algorithms improves execution time and completion time of the last task, as well as energy consumption by 18.30%, 15.14%, and 10.21%, respectively, compared to the semi-greedy method.

The network traffic is increasing daily; consequently, software-defined network technology is employed to manage the network, as this technology provides an overview of the network and enables advanced management. In software-defined networks, load balancing is also necessary to improve performance. Many approaches have been proposed for load balancing in software-defined networks. These approaches can be the taxonomy; however, the taxonomies presented so far are not exact. In this article, a detailed taxonomy for load balancing of software-defined networks is provided. Then, the approaches that use optimization algorithms based on artificial intelligence to address the load balancing challenge in software-defined networks are discussed. Finally, the methods for predicting the load balance in software-defined networks and how this contributes to reducing energy consumption are presented.

In Iran, the stock market is facing different conditions compared to the rest of the world. One of the most important challenges in this market is the lack of transparency in market information and information on trading companies. Also, the lack of appropriate and complete historical data for use in forecasting algorithms is another important challenge. In stock price forecasting, due to the dynamic interactions of the stock market and price changes in short periods of time, the use of artificial intelligence is used as a powerful tool in price forecasting and decisions related to buying and selling stocks. In this paper, a method based on machine learning, including five steps (including data labeling, feature extraction, feature selection, classification, and signal presentation), is presented. For this purpose, various technical characteristics have been extracted from the price data, and the data has been labeled using the threshold labeling method. Then, various machine learning models are trained on this data and provide buy and sell signals at the output. To improve the performance of the machine learning model, feature selection has been done using the Cuckoo Search algorithm. In order to evaluate the proposed method, several years of Iranian stock market data and various indices have been used. The results of the evaluation show the effectiveness of the proposed method.

The congestion of roads is a very important factor in urban traffic. A lot of research has tried to solve these problems using meta-heuristic algorithms. In these algorithms, firstly, routing is done randomly over large areas. This will increase the search time. In addition, these algorithms only consider the physical distance between the vehicles. Since environmental factors such as traffic are very effective in routing, these factors should be considered in routing. In this paper, to solve the problems, a dynamic path programming method based on the combination of the ant colony algorithm and particle swarm optimization, along with a cosine function of angle, has been proposed. This method takes into account various factors of roads, such as the length of the urban road and the incoming and outgoing traffic at intersections. In the method, the points that are aligned with the navigation path towards the final destination are given more chances. The results of applying the proposed model on the valid data of the TSPLIB library, which is based on the physical distance between cars, show that the search time of the proposed method has decreased by 40.74% on average compared to the results of ten other methods used for evaluation. The highest and lowest rates of decrease are 98.01% and 6.02%, respectively. The test of dynamic route planning under road traffic on some intersections of Beijing city also shows that the proposed method only causes congestion of about 1.57%.