·         Table of Contents (Volume #8, Issue #4)

·         Cover Page (Volume #8, Issue #4)

·         Editorial (Volume #8, Issue #4)

Paper #1

Trans-Compiler-Based Conversion from Cross-Platform Applications to Native Applications

Amira T. Mahmoud, Moataz-bellah Radwan, Abdelrahman Mohamed Soliman, Ahmed H. Youssef, Hala H. Zayed and Walaa Medhat

Abstract: Cross-platform mobile application development is emerging widely in the mobile applications industry. Cross-platform Frameworks (CPFs) like React Native, Flutter, and Xamarin are used by many developing companies. The technology these frameworks use faces performance and resource use efficiency limitations compared to native applications. The native applications are written in the native languages of the platforms. Trans-complier-based conversion between native languages of different platforms of mobile applications has been addressed in recent research. However, the problem statement needed to be mathematically represented. The solution depended on hard coding and needed more generalization. In addition, it might not be a practical solution for companies that are using and already have built applications using CPFs. Therefore, in this paper, we present an enhanced-trans-compiler-based converter to convert applications made by CPFs to native applications. We implemented the architecture to convert React Native and Xamarin applications. The React Native to Native tool converted thirteen applications to native Android and iOS applications, with accuracies ranging from 40% for large applications to 100% for simple applications. The maximum conversion time was seven minutes for converting 40% of an 8K LOC application. In addition, since Large Language Models (LLMs) are the trendiest technology in our era, we compared our proposed solution output with LLMs. We concluded its superiority compared to the status of LLMs. Performance evaluation is also done to compare the React Native applications against native applications generated by the trans-compiler tool. The assessment showed that the native applications perform better than React Native regarding runtime memory consumption, storage, and speed. The Xamarin to Native tool was also tested to show the genericness of the architecture and how it can be extended to convert from any CPF to Native applications.

Keywords: Code generation; Cross-platform; Mobile applications; Innovation; Resource use efficiency; Trans-compilation.

Download Full Text

Paper #2

A Systematic Literature Review on Cross Domain Sentiment Analysis Techniques: PRISMA Approach

Rati Sharma and Kamlesh Lakhwani

Abstract: Cross Domain Sentiment Analysis (CDSA) is a method that uses rich and quality-labeled source domain data to identify the sentiments of poorly or without labeled target data. In the past decade, ample research studies have focused on this topic to solve and propose efficient CDSA methods. However, an extensive investigation of these past studies is required to find a window of improvement. The main aim of the study is to figure out considerable developments, methodologies, and SOTA techniques in the recent past. This research study presents a systematic literature review to analyze the CDSA studies published from 2017 to 2023. The authors have selected 34 articles overall and categorized them into seven different CSDA techniques. The extensive analysis of these studies’ results (in the form of graphs and tables) into different parameters that impact the performance of the CDSA. The survey finds out that major research studies tried to create a relationship between pivots and non-pivots to gain accuracy. This paper presents a synthesized review of CDSA and explores the current methods and potential future directions. It also addresses the challenges and opportunities presented by these emerging trends and their significance for researchers and practitioners in the CDSA field.

Keywords: Cross Domain Sentiment Analysis; Domain Adaptation; PRISMA 2020; Transfer Learning.

Download Full Text

Paper #3

Variance Consistency Learning: Enhancing Cross-Modal Knowledge Distillation for Remote Sensing Image Classification

Huaxiang Song, Yong Zhou, Wanbo Liu, Di Zhao, Qun Liu and Jinling Liu

Abstract: Vision Transformers (ViTs) have demonstrated exceptional accuracy in classifying remote sensing images (RSIs). However, existing knowledge distillation (KD) methods for transferring representations from a large ViT to a more compact Convolutional Neural Network (CNN) have proven ineffective. This limitation significantly hampers the remarkable generalization capability of ViTs during deployment due to their substantial size. Contrary to common beliefs, we argue that domain discrepancies along with the RSI inherent natures constrain the effectiveness and efficiency of cross-modal knowledge transfer. Consequently, we propose a novel Variance Consistency Learning (VCL) strategy to enhance the efficiency of the cross-modal KD process, implemented through a plug-and-plug module within a ViTteachingCNN pipeline. We evaluated our student model, termed VCL-Net, on three RSI datasets. The results reveal that VCL-Net exhibits superior accuracy and a more compact size compared to 33 other state-of-the-art methods published in the past three years. Specifically, VCL-Net surpasses other KD-based methods with a maximum improvement in accuracy of 22% across different datasets. Furthermore, the visualization analysis of model activations reveals that VCL-Net has learned long-range dependencies of features from the ViT teacher. Moreover, the ablation experiments suggest that our method has reduced the time costs in the KD process by at least 75%. Therefore, our study offers a more effective and efficient approach for cross-modal knowledge transfer when addressing domain discrepancies.

Keywords: Cross-Modal; Deep Learning; Knowledge Distillation; Remote Sensing Image Classification.

Download Full Text

Paper #4

Neural-Based Secured Decentralized E-Voting Framework using Blur Image Broadcasting

Samayamanthula Venkata Chinnaiah Gupta and Kodati Satya Prasad

Abstract: One of the fundamental rights in the modern democracy is voting. Much research has been done to strengthen the voting process and security. The new and safe Neural-Based Secured Decentralised E-Voting Framework employing Blur Image Broadcasting tackles the major issues with standard electronic voting techniques. Neural networks with blurred image broadcasting protected voter confidentiality, ballot integrity, & system security. Therefore, a novel Zebra-based GoogleNet Elliptic Curve (ZbGEC) is provided to upgrade the decentralized e-voting via blur image broadcasting in this study. It authenticates and broadcasts the voter’s information safely to the blockchain technology. It additionally demonstrates time consumption, memory usage, and design cost. Only authorized users can view and alter encrypted and decrypted votes via neural networks. This encryption protects voter anonymity and ballot manipulation. Furthermore, blur image streaming obscures voter ballot selections, improving voter privacy. The decentralized design spreads voting over numerous nodes; removing the centralized Spread structure strengthens the system against manipulation and cyber-attacks. Notably, decentralized e-voting time consumption and response time were minimized to an efficient 2 seconds and 5 seconds. The proposed system's design cost was economical at $30, while memory usage was optimized to 300 MB, representing a significant improvement over traditional methods. Neural-based security, decentralized structures, and blurred image streaming produce a reliable e-voting system. This architecture improves security, privacy, openness, and scalability over electronic voting systems. The Neural-Based Secured Decentralised E-Voting Framework utilizing Blur Image Broadcasting might make voting safer, more transparent, and inclusive.

Keywords: Blockchain Security; Blur Image Broadcasting; Decentralized E-Voting Framework; Neural-Based.

Download Full Text

Paper #5

Whale-Based Trajectory Optimization Algorithm for 6 DOF Robotic Arm

Mahmoud A. A. Mousa, Abdelrahman T. Elgohr and Hatem A. Khater

Abstract: Trajectory optimal control for a robotic arm with a high degree of freedom (DOF) is challenging. The design space for that problem is complex and the search for an optimal solution is demanding. The design of a robotic arm's trajectory is based on solving the inverse kinematics problem, considering additional refinements influenced by factors like total rotating angle, reachability time, minimum execution time, obstacle avoidance, and energy consumption minimization. Due to the complexity of the design space, in this paper, genetic algorithm (GA) optimization and whale optimization algorithm (WOA) have been used to achieve robotic arm trajectory control while maintaining a minimum reachability time. To validate the suggested techniques, a case study was conducted on a 6 DOF KUKA KR 4 R600 robot arm to control subject to its constraints. Sets of consecutive points forming four different paths were inputted to the algorithms. The goal was to reach all these points, in order, with a minimum total reachability time. As a result of this paper, we shown that the whale optimization algorithm provides better performance than the genetic algorithm with a factor of more than 2.5 while satisfying the reachability constraints.

Keywords: Genetic Algorithm; Kinematics Analysis; Trajectory Optimization; Whale Optimization Algorithm; 6 DOF Robotic Arm.

Download Full Text