Daily Archives: June 30, 2026

Authors: Lavish Kumar, Mohd Aamish, Murad Aalam, Himanshu Kumar Thakur, Ashwani Dubey, Dr. Raj Kumar

Abstract: The use of artificial intelligence (AI) technology for medical image analysis has gained significant importance in contemporary health care systems. AI models assist physicians in diagnosing diseases based on medical images like x-rays, MRIs, CT scans, and ultrasound scans with great precision and fast diagnosis. Nonetheless, medical datasets involve critical and sensitive data about patients, thus raising serious concerns regarding privacy protection in the context of AI applications. The exposure or unauthorized access of medical data can result in severe ethical and legal problems [2], [9], [16]. In this paper, we present a privacy-aware medical image analysis system based on the implementation of convolutional neural networks (CNN), PyTorch framework, Streamlit toolkit, and Differential Privacy technology. CNN is used to extract the features of the medical images automatically and classify the underlying diseases. PyTorch is used for developing the proposed model efficiently, and Streamlit provides a user-friendly interface for physicians. Moreover, the training process is implemented based on differential privacy in order to maintain the privacy of the data. [4], [5]. The proposed framework can support hospitals, diagnostic centers, and telemedicine systems in secure healthcare applications [6], [20].

DOI: http://doi.org/10.5281/zenodo.21063384

Published by: vikaspatanker

Authors: Anjali Kapoor, Dr Mridula

Abstract: In order to facilitate dynamic customization in contemporary learning environments, this study introduces a Smart Autonomous System for Adaptive Student Profiling. To create constantly changing student profiles, the suggested architecture incorporates a variety of educational data sources, such as academic achievement, behavioral patterns, cognitive traits, environmental data, social interactions, and emotional indications. To handle missing data, the system uses KNN-based imputation Convolutional Neural Networks (CNNs) for emotion identification, Principal Component Analysis (PCA) for feature reduction, and XG Boost for academic risk prediction and student profile. A Deep Q-Network (DQN)-based reinforcement learning method that modifies suggestions and interventions based on learner needs enables autonomous decision-making. A hybrid recommendation engine also facilitates optimum learning paths and individualized material distribution. Real-time profiling, ongoing monitoring, proactive intervention, and adaptive feedback mechanisms are made possible by the framework's implementation within a multi-agent architecture. Learner engagement, academic achievement, and early detection of at-risk kids have all improved, according to experimental evaluation utilizing benchmark educational datasets. The findings show that the suggested approach is reliable, scalable, and successful in facilitating intelligent, customized learning environments.

DOI: https://doi.org/10.5281/zenodo.21063202

Published by: vikaspatanker

Authors: Mr. Kuldeep, Associate Professor Dr. Pramod Kumar

Abstract: The Industrial Internet of Things (IIoT) has transformed industrial ecosystems by enabling real-time monitoring, automation, and data-driven decision-making. Deep learning techniques have emerged as powerful tools for predictive analytics, supporting applications such as anomaly detection, fault diagnosis, and predictive maintenance. However, centralized deep learning approaches introduce significant security and privacy risks, including data leakage, adversarial attacks, and model poisoning. This research proposes a secure hybrid deep learning framework integrating CNN-LSTM with ANFIS, along with federated learning, blockchain, and differential privacy to ensure secure, privacy-preserving, and explainable predictive analytics in IIoT environments. The framework enhances prediction accuracy while maintaining data confidentiality, robustness, and real-time performance.

DOI: https://doi.org/10.5281/zenodo.21060033

Published by: vikaspatanker

Authors: Sk Samsul Hoda, Professor Dr Vipin kumar

Abstract: Bending and focusing magnets, both normal- or super-conducting, are crucial elements for the performance of any particle accelerator. Their design require-ments are always more tighten regarding components’ misalignment and mag-netic properties. This dissertation proposes new solutions for characterizing mag-netic materials and monitoring solenoids’ magnetic axis misalignments. A superconducting permeameter is designed to characterize the new-generation superconducting magnet yokes at their operational temperature and saturation level. As proof of principle, the magnetic characterization of ARMCO®Pure Iron was performed at the cryogenic temperature of 4.2 K and a saturation level of nearly 3 T. A case study based on the new HL-LHC superconducting magnets quantifies the impact of the magnetic properties of the yoke on the performances of the superconducting magnets. A flux-metric based method is proposed to identify the relative magnetic perme-ability of weakly magnetic materials. As proof of principle, the magnetic prop-erties of the ITER TF coils quench detection stainless steel are analyzed. This method is not suitable to test materials with a relative permeability lower than 1.1. Hence, a measurement system based on a new magneto-metric method is conceived and validated employing a standard reference sample. The methods proposed in this thesis are currently employed at CERN’s magnetic laboratory to face an increasing number of requests concerning not only the magnetic charac-terization of materials for magnets but also for shielding systems and compatibil-ity of various components with high magnetic fields. In this thesis, the results of the evaluation of ARMCO®Pure Iron as the yoke of the new LHC superconducting magnets and CRYOPHY as the magnetic shield for the cryomodule prototypes of HL-LHC Crab Cavities are reported. Finally, a new Hall-sensor method is conceived and implemented for monitor-ing the coils alignment in multi-coil magnets, directly during their operation in particle accelerators. The proposed method is suitable even for those cases when almost the whole magnet aperture is not accessible. Requiring only a few mea-surements of the magnetic field at fixed positions inside the magnet aperture, the method overcomes the main drawback of the other Hall sensor-based methods which is having to deal with sturdy mechanics of the moving stages. The method is validated numerically on a challenging case study related to the Solenoid B of the project ELI-NP.

DOI: https://doi.org/10.5281/zenodo.21058750

Published by: vikaspatanker