This paper proposes an interpretable cross-modal neural architecture for fusing time-synchronized data across complex workflows. Leveraging temporal attention and visual-textual embeddings, the model enhances real-time analytics by improving semantic consistency and reducing fusion ambiguity in multimodal environments.

Rohan Prakash Menon, Olivia Carter, Jean Philippe Moreau, Arvind Ramesh, Taro Masahiro Suzuki

We introduce a reinforcement learning framework for autonomous energy optimization in smart grid-edge systems. The model dynamically adjusts consumption strategies based on predictive analytics, improving load balancing, reducing energy waste, and enabling scalable deployment across heterogeneous infrastructural layers.

Liu Zhen Wei, Chen Hao Ming, Xu Li Yuan, Huang Xiang Lei, Wang Qiang Bo

This study surveys and implements self-supervised learning strategies to address challenges in high-dimensional visual annotation. Utilizing contrastive learning, autoencoding, and attention mechanisms, the proposed pipeline reduces reliance on labeled datasets while boosting annotation accuracy and generalizability in real-world deployments.

Amit Deepak Kulkarni, Lucia Isabella Romero, Kunal Rajiv Desai, Ivan Andrei Petrov, Priya Chandrika Sinha

We propose a transformer-augmented graph neural network for scalable anomaly detection in streaming IoT networks. The model captures contextual dependencies through spatial-temporal embeddings, improving detection precision and enabling dynamic adaptation in evolving sensor environments under limited supervision.

Gao Wen Jie, Zhang Yu Chen, Liu Xiao Long, Chen Rong Ming, Huang Liang Zhe

This paper introduces a hybrid reasoning system combining symbolic logic with neural computation for critical infrastructure monitoring. It ensures transparent decision-making, traceability of predictions, and real-time alert generation for anomalies in mission-critical systems using interpretable rule-based logic and contextual neural analysis.

Robert James Whitman, Emily Charlotte Spencer, Thomas Gregory Adams, Laura Catherine Bennett, Michael Alexander Foster

This paper presents a federated learning framework integrated with secure aggregation to protect data privacy in distributed institutional learning. The approach ensures collaborative model training without raw data sharing while maintaining accuracy across institutions with varying data sizes and feature distributions.

Sarah Elizabeth Thompson, Emma Rose Walters, Benjamin Charles Griffin, Olivia Marie Dean, Thomas Henry Williams