Recent Trends in Artificial Intelligence & It’s Applications (e-ISSN: 2583-4819)

Quantum Circuits for Deep Neural Networks: Potential, Pitfalls, and Progress

D. Sri Varshini, S. Geetha, Chandra Sekhar Koppireddy — Thu, 11 Sep 2025 00:00:00 +0000

The conjunction of deep learning and quantum computing marks a turning point in the history of intelligent systems. Deep learning has transformed Artificial Intelligence (AI) by enabling machines to learn sophisticated representations from large datasets, greatly impacting image recognition, natural language processing, and autonomous systems. However, training deep neural networks requires immense computational power, often limited by traditional hardware constraints. Quantum computing, which relies on superposition, entanglement, and quantum parallelism, offers a new paradigm that could overcome some of these computational limitations. This chapter explores the rapidly evolving intersection of deep learning and quantum computing, known as Quantum Deep Learning (QDL) or Quantum Machine Learning (QML). We begin by introducing fundamental concepts from both fields, followed by a comprehensive overview of cutting-edge quantum-enhanced learning methods, such as quantum neural networks, variational quantum circuits, and hybrid quantum-classical architectures. Particular focus is given to how quantum algorithms might accelerate deep learning processes, improve model generalization, and enable new forms of learning representations for quantum data. Despite these promising prospects, significant challenges remain. Current quantum hardware, mostly in the Noisy Intermediate-Scale Quantum (NISQ) stage, suffers from qubit decoherence, gate noise, and limited scalability. Moreover, encoding classical information into quantum states and optimizing parameterized quantum circuits are complex tasks. This chapter examines these challenges and discusses the theoretical and practical hurdles that must be overcome to realize practical quantum advantage in deep learning. It concludes with open research questions, potential application areas, and a future research roadmap at this exciting interdisciplinary frontier.

Transforming Higher Education: The Integral Role of AI in Learning Outcomes and Integration across Disciplines

Shailja Gaur, Sunny Gaur — Tue, 14 Oct 2025 00:00:00 +0000

The subject of this paper considers how artificial intelligence (AI) might be used in higher education to help students maintain a more equitable learning environment. While a lot of media coverage portrays AI as a danger to higher education, we argue that technology may be efficiently employed to build an additional equitable and inclusive learning environment rather than helping students overcome learning obstacles or improve accessibility. To ensure that students utilize AI responsibly and follow their institution's academic integrity regulations, the authors also offer helpful recommendations for integrating AI into teaching, learning, and evaluation. However, maintaining academic integrity requires citing AI sources, and students require teacher supervision and direction. The use of AI tools, deep learning, GAI in context with teaching and learning is examined in the content. It also looks on learning analytics, AI applications for automated grading, virtual assistants, and personalized learning. The study concluded by urging educators to interact with AI, talk to students about it, and explore its potential and drawbacks in teaching in their field. To improve learning outcomes, foster creativity, and carefully assess the potential effects of AI in the classroom, needs to acquire academic policies and practices that support educational equity and quality.

Phishing Attack Mitigation Through AI: A Review of Feature Engineering and Classification Techniques with Model Interpretability

Nayan Pundlik, Tripti Saxena — Thu, 11 Sep 2025 00:00:00 +0000

Phishing, which targets unwary users with fake emails, URLs, and websites to steal sensitive data, remains one of the most sophisticated and persistent cybersecurity threats. Phishing attacks are becoming more complex and frequent, making traditional rule-based protection strategies inadequate. Artificial Intelligence (AI) has emerged as a powerful alternative, providing adaptive, scalable, and precise detection methods. A comprehensive review of AI-driven phishing mitigation techniques emphasizes three key areas: feature engineering, classification methods, and model interpretability. The study examines commonly used features, including URL-based, domain-based, content-based, and behavioral features, along with advanced feature selection and dimensionality reduction strategies such as PCA, LDA, and autoencoders. It discusses both traditional machine learning classifiers SVM, Random Forest, k-NN and deep learning architectures like CNN, RNN, and LSTM, as well as ensemble techniques such as XGBoost and AdaBoost. Recognizing the complexity of AI models, it highlights interpretability techniques like SHAP, LIME, and inherent model transparency to improve trustworthiness and explainability in phishing detection.