Technological advancements in deep-learning based models have resulted in the utilization of sophisticated architectures to model discourse in dialogue systems. However, despite their effectiveness, these systems lack transparency in the discourse modeling process, making them difficult to adapt and interpret. Recently, spectral-learning based algorithms have gained interest in data-driven approaches. This study proposes two spectral algorithms using canonical correlation analysis to develop two discourse modeling frameworks for a multi-turn retrieval-based dialogue system. The proposed spectral-learning based models learn discourse units that capture the intent/hidden associations in a conversation. Both models utilize the multi-view nature of textual data, i.e. each turn in a turn-taking conversation. The proposed variants aim to construct an attention state: the first develops a global attention state consisting of global discourse units abstracted to encapsulate the long-term dependencies in a conversation. On the other hand, the second model establishes a local attention state consisting of local discourse units conceptualized as the focus of attention for each utterance in a conversation. Subsequently, the models employ Word Movers Distance to measure the semantic distance between the established attention state and the candidate dialogues to retrieve the top-k-ranked candidate responses. The models are simple and adaptable, eliminating the need for a large labeled corpus. Qualitative and quantitative evaluations on the UBUNTU dataset demonstrate the efficacy of the proposed models.

The complex industrial system, which is subject to different influencing factors, often manifests itself in several failure modes, making the use of different standard and unimodal distributions to model its behavior unnecessary and inappropriate. In this article, we are interested in presenting an approach to modeling the dynamic behavior of the system based on a Mixture Weibull Proportional Hazard Model. In addition to the advantage of proportional hazard models taking into account the influence of covariates on system behavior, the use of mixed Weibull models makes it possible, on the basis of a mixing parameter W_i, to highlight the weight of each component i on the overall and dynamic behavior of the system. This approach is illustrated first by considering data generated by the MATLAB programming language by justifying the contribution that can be obtained by this mixture model in the modeling of the reliability of complex systems. Then, from a history of maintenance and reliability of a gas turbine having operated for more than thirteen years in the SONATRACH company, the maximum likelihood approach and the likelihood ratio test makes it possible to validate the goodness of fit of the proposed model and to estimate the influence in the probabilities of failures of two heterogeneous subpopulations representing hidden behaviors. Then, the mixed Weibull model will be extended to incorporate other covariates of the system by constructing the proportional hazard model (PHM). The proposed model is validated by the Akaike Information Criterion (AIC) and the Bayes information criterion (BIC) based on the maximum likelihood value. This approach facilitates decision-making on system intervention, taking into account operating conditions and prioritizing the most critical subsystem over time.

In recent years, researchers have shifted their focus from effective security to technologies that enhance service quality. For a resource-constrained system such as the Internet of Things (IoT), there must be a trade-off between service quality and security. Access control is the most important responsibility for any system, and in this area of the system, we may increase security by using blockchain. Resource-constrained systems, due to their limited memory and small settings, cannot utilize blockchain technology. The problem with blockchain is that resource-constrained systems cannot handle its intricate and time-consuming calculations. If we can balance the complexity and viability of blockchain, we can easily implement it for security solutions. We can make blockchain lighter by modifying the Proof of Work (PoW), the overall framework, or the hash function. To make blockchain more lightweight, one can use several quicker hash coding methods. In this study, we examine various hashing techniques and employ blockchain technology to evaluate the results. This article provides a numerical analysis to evaluate the efficacy of various hashing strategies on blockchain, enabling their application in resource-constrained systems such as the Internet of Things, electronic health record systems, and voting systems, among others. In our work, we provide a novel mathematical analysis of the hash function in a blockchain system. We analyzed the suitability of a hash function for resource-constrained devices using Python. Experimental results show that time consumption is as low as 26-29% at various difficulty levels, demonstrating the feasibility of the BLAKE2s() hash function as compared to the traditional SHA256() hash function on blockchain. This work aims to establish a solid foundation for researchers and designers of resource-constrained systems seeking to integrate blockchain for enhanced security in next-generation devices.

Software fault prediction plays a crucial role in maintaining software quality by identifying modules that are prone to defects early in the development cycle. Nevertheless, issues including class imbalance, high-dimensional data, and the shortcomings of individual classifiers make prediction models less successful. This paper addresses these issues by putting up a new Balanced Voting-PCA Classifier with SMOTE-ENN (BVPC-SENN) model. The BVPC-SENN model incorporates a weighted voting ensemble of Bernoulli Naive Bayes (BNB), Gaussian Naive Bayes (GNB), Random Forest (RF), and Support Vector Machines (SVM) as base classifiers, to handle class imbalance, and Principal Component Analysis (PCA) for dimensionality reduction. In order to provide reliable and accurate fault predictions, the BVPC-SENN model balances the dataset, reduces feature dimensions, and combines the predictions of many classifiers using a weighted voting method. Experiments on several benchmark datasets show that the BVPC-SENN model achieves improved accuracy, precision, and generalization, greatly enhancing prediction performance. The suggested methodology improves the state-of-the-art in software fault prediction and provides a useful framework for strengthening software quality assurance procedures by successfully addressing class imbalance, optimizing feature representation, and utilizing ensemble learning.

Efficient scheduling in industrial production remains a critical challenge, requiring innovative approaches to address modern manufacturing systems' growing complexity and dynamic nature. This paper presents a hybrid optimization approach combining genetic algorithm (GA) and multi-agent systems (MAS) to tackle the challenges of machine scheduling in complex production environments. The study highlights significant advancements in reducing downtime, idle time, and production time, improving efficiency and resource utilization. GA provides optimal task sequencing and scheduling, while MAS enables real-time collaboration and dynamic adjustments of maintenance schedules, ensuring system adaptability to operational changes. The results underscore the robustness and versatility of the GA-MAS strategy, offering a practical and innovative solution to enhance productivity in modern industrial systems.

This research introduces a new way to detect and predict crime using modern CCTV cameras. It uses a “software as a service” model to make crime monitoring and analysis more affordable. The system works with both old and new cameras, even those without GPS, by using latitude and longitude mapping. This means it can detect crimes in areas with poor internet connection. The method ensures data remains private and secure, which is important due to the sensitive nature of crime information. The paper also explains how to install the software on different types of cameras. Special software for police officers provides instant crime updates, helping them respond faster and more effectively. The study shows that analyzing crime data can identify high-risk areas, allowing authorities to prevent crimes before they happen. Overall, this approach is promising for helping police and investigators reduce crime and improve public safety. It also opens up new opportunities for research in crime detection technology. The study explores various deep learning methods for image recognition and suggests a real-time alert system for law enforcement, using tools like TensorFlow, Google Maps, and Firebase. It highlights the importance of involving people, using smart video analysis, and advancing technologies like computer vision, federated learning, and edge computing to improve crime detection.