Zahid Iqbal

Automated silicon wafer defect inspection is critical for sustaining yield and reliability in semiconductor manufacturing. Traditional image processing and feature engineering approaches struggle with complex and evolving defect patterns, while deep learning methods, although effective, increase computational cost and risk of overfitting. Many existing models address class imbalance through synthetic data, which can introduce artifacts and misrepresent the actual defect distributions. Additionally, conventional convolutional neural networks remain limited in capturing global context, leading to the misclassification of visually similar defects. This study addresses these issues through a simple yet effective two-stage approach. The proposed ensemble attention convolutional neural network (EACNN) first performs binary defect detection to distinguish between defective and non-defective samples, followed by a dedicated defect-type classification for fine-grained categorization. This staged design decomposes the complex multi-class problem into manageable subtasks and mitigates class dominance without synthetic data. The model further incorporates spatial and channel attention mechanisms to strengthen discriminative feature learning and improve the representation of salient defect regions. Focal loss is employed to mitigate class imbalance by down-weighting easily classified examples and placing greater emphasis on hard, misclassified samples. Optimization strategies, including batch normalization, dropout, and global average pooling, are adopted to improve generalization, reduce parameter complexity, and accelerate convergence. Experiments on the WM-811K dataset through 10-fold cross-validation demonstrated the effectiveness of EACNN and achieved 95.57% accuracy, 93.67% precision, 95.41% recall, 98.76% specificity, 94.47% F-measure, and 94.51% G-Mean. Comparisons with classical, attention-based, and imbalance-handling methods confirmed the superior performance of EACNN. The results underscore EACNN’s ability to learn highly discriminative features while preserving the integrity of real defect distributions.

Because of the evolution of artificial intelligence (AI), home security has progressed from a basic security system to an active architecture that is responsive and adaptive to real world situations. Due to the rapid adoption of AI in smart systems, there is increasing suspicion surrounding privacy issues and ethical ambiguity, as well as gaps when it comes to regulating these technologies. We provide an overview of AI in smart home security applications and examine the area of security, access control, intrusion detection, human action recognition, and research on intelligent automation. We summarize the last decade of evolution, with some summaries of previous on computer vision, authentication systems, and finding unusual patterns recently. Our key findings include the development of approaches to improve real time security monitoring, dramatic reductions in false alarms, and customization of home access using AI. Improvements in security have also increased risk with respect to ethical ambiguity as well as technical issues in certain cases. In this paper, we offers pathways for improved AI system design, proposed formal data protection regulations, and examples of simplifying complex system for user comprehension, which also establishes the groundwork for future efforts. Home security should balance new opportunities with ethical considerations.

Blockchain has emerged as a groundbreaking security technology, playing a vital role in various industries such as banking, the Internet of Things (IoT), healthcare, education, and voting. However, the widespread adoption of this technology has introduced certain vulnerabilities, particularly in the form of exploitation by malicious entities. While existing research primarily focuses on identifying anomalous actor behavior, there has been limited exploration of precisely identifying hostile actors within the Ethereum network. This study aims to uncover malevolent actors operating on the Ethereum network and categorize attacks based on their actions. To achieve this research goal, a new dataset was constructed by consolidating data on malicious actors involved in illicit Ethereum activities. Key features were extracted from this dataset using advanced feature selection techniques, including Principal Component Analysis (PCA), Information Gain, and Ridge Regression. Machine learning classifiers such as LGBM, XGBoost, Random Forest, Extra Tree, Bagging, and K-Nearest Neighbors were applied to identify and classify malicious actors effectively. The results, achieving an impressive accuracy rate of 98%, underscore the effectiveness of Information Gain when coupled with LGBM and XGBoost. Notably, XGBoost demonstrates efficiency by completing the analysis in a mere 13.72 seconds. In addition to identifying fraudulent activities, this research classifies them into distinct categories, enhancing blockchain security and addressing trust concerns. This study’s outcomes fortify the Ethereum network’s resilience and contribute to the broader discourse on bolstering reliability in blockchain systems.

This paper addresses the problem of decentralized learning to achieve a high-performance global model by asking a group of clients to share local models pre-trained with their own data resources. We are particularly interested in a specific case where both the client model architectures and data distributions are diverse, which makes it nontrivial to adopt conventional approaches such as Federated Learning and network co-distillation. To this end, we propose a new decentralized learning method called Decentralized Learning via Adaptive Distillation (DLAD). Given a collection of client models and a large number of unlabeled distillation samples, the proposed DLAD 1) aggregates the outputs of the client models while adaptively emphasizing those with higher confidence in given distillation samples and 2) trains the global model to imitate the aggregated outputs. Our extensive experimental evaluation on multiple public datasets (MNIST, CIFAR-10, and CINIC-10) demonstrates the effectiveness of the proposed method.

New technologies are adapted to made progress in healthcare especially for independent livings. Medication at distance is leading to integrate technologies with medical. Machine learning methods in collaboration with wearable sensor network technology are used to find hidden patterns in data, detect patient movements, observe habits of patient, analyze clinical data of patient, find intention of patients and make decision on the bases of gathered data. This research performs comparative study on non-clinical systems in healthcare for independent livings. In this study, these systems are sub-divided w.r.t their working into two types: single purpose systems and multi-purpose systems. Systems that are built for single specific purpose (e.g. detect fall, detect emergent state of chronic disease patient) and cannot support healthcare generically are known as single purpose systems, where multi-purpose systems are built to serve for multiple problems (e.g. heart attack etc.) by using single system. This study analyzes usages of machine learning techniques in healthcare systems for independent livings. Answer Set Programming (ASP), Artificial Neural Networks, Classification, Sampling and Rule Based Reasoning etc. are some state of art techniques used to determine emergent situations and observe changes in patient data. Among all methods, ASP logic is used most widely, it is due to its feature to deal with incomplete data. It is also observed that system using ANN shows better accuracy than other systems. It is observed that most of the systems created are for single purpose. In this work, 10 single purpose systems and 5 multi-purpose systems are studied. There is need to create more generic systems that can be used for patients with multiple diseases. Also most of the systems created are prototypical. There is need to create systems that can serve healthcare services in real world.

Assurance of learning is the key to students’ academic success. To this end, one of the largest historically Black colleges and universities (HBCUs) in the United States administers the Educational Testing Service Major Field Test for Business (ETS MFT-B) to assess student learning of business concepts. This study used a set of predictor variables to explain the ETS MFT-B performance at the subject HBCU. The results of the study are expected to close the gap between student performance and expected outcomes in the subject institution, as well as contribute to the broader issue of improving curriculum and teaching effectiveness in business schools in the United States. Using data from fall 2012 to fall 2018, the findings of the regressions indicated that new variables, namely, English language proficiency, hours attempted to hours earned, first-generation student status, and desire to pursue higher studies were important predictors of the ETS MFT-B total score and sub-scores. Among the traditional variables, the cumulative GPA and student’s major area were important predictors. These findings point toward several corrective actions in the assurance-of-learning process for the subject school. The subject school can add a business English class to the curriculum, improve teaching effectiveness, provide advising assistance when students retake classes, and adopt measures that will motivate students to prepare for the ETS MFT-B.

New technologies are being adapted to made progress in healthcare especially for independent livings. Medication at distance is leading to integrate technologies with medical. Machine learning methods in collaboration with wearable sensor network technology are used to find hidden patterns in data, detect patient movements, observe habits of patient, analyze clinical data of patient, find intention of patients and make decision on the bases of gathered data. This research performs comparative study on non-clinical systems in healthcare for independent livings. In this study, these systems are sub-divided w.r.t their working into two types: single purpose systems and multi-purpose systems. Systems that are built for single specific purpose (e.g. detect fall, detect emergent state of chronic disease patient) and cannot support healthcare generically are known as single purpose systems, where multi-purpose systems are built to serve for multiple problems (e.g. heart attack, fall detection etc.) by using single system. This study analyzes usages of machine learning techniques in healthcare systems for independent livings. Answer Set Programming (ASP), Artificial Neural Networks, Classification, Sampling and Rule Based Reasoning etc. are some state of art techniques used to determine emergent situations and observe changes in patient data. Among all methods, ASP logic is used most widely, it is due to its feature to deal with incomplete data. It is also observed that system using ANN shows better accuracy than other systems. It is observed that most of the systems created are for single purpose. In this work, 10 single purpose systems and 5 multi-purpose systems are studied. There is need to create more generic systems that can be used for patients with multiple diseases. Furthermore, most of the systems created are prototypical. There is need to create systems that can serve healthcare services in real world. Some systems are hard to be used in real life due to large hardware requirements. Although, good systems are created but still there is need to build more efficient, affordable, adoptive and generic systems.

In real life, mostly problems are dynamic. Many algorithms have been proposed to handle the static problems but these algorithms do not handle or poorly handle the dynamic environment problems. Although, many algorithms have been proposed to handle dynamic problems but still, there are some limitations or drawbacks in every algorithm regarding diversity of particles and tracking of already found optima. To overcome these limitations/drawbacks, we have proposed a new efficient algorithm to handle the dynamic environment effectively by tracking and locating multiple optima and also by improving the diversity and conversion speed of algorithm. In this algorithm, a new method has been proposed which explore the undiscovered areas of search space to increase the diversity of algorithm. This algorithm also uses a method to effectively handle the overlapped and overcrowded particles. Branke has proposed moving peak benchmark which is commonly used MBP in literature. We also have performed different experiments on Moving Peak Benchmark. After comparing the experimental results with different state of art algorithms, it was seen that our algorithm performed more efficiently.