This research presents a federated learning framework with attention-based aggregation for predictive maintenance. The system supports multi-vendor collaboration while preserving data privacy, enhancing fault prediction accuracy and reducing unplanned downtimes in industrial Internet of Things infrastructures.

Ravi Shankar Kulkarni, Maria Isabella Lopez, Aditya Suresh Reddy, Jean Pierre Roux, Taro Kenji Yamamoto

We propose a contrastive learning approach using graph matching for event detection in surveillance streams. The model aligns visual and audio data, enhancing detection performance across crowded and acoustically complex public spaces using few labeled events.

Chen Jia Hao, Liu Tian Cheng, Xu Zhen Wei, Gao Xiao Lin, Zhang Min Hui

This paper introduces a spatiotemporal energy forecasting model leveraging multi-head encoders and attention regularization. The system accurately predicts weather-driven power demand fluctuations, aiding grid operators in managing peak loads and renewable energy transitions in urban zones.

Michael Scott Wilson, Laura Jane Bennett, Olivia Claire Grant, Emily Renee Howard, Nathan Gregory Ellis

This study introduces a federated learning method for personalized healthcare prediction. The model integrates adaptive aggregation and cross-institutional model sharing to ensure privacy preservation while enhancing accuracy and generalizability in multi-site clinical environments with heterogeneous patient data.

Amit Raghav Sharma, Fatima Noor El-Sayed, Taro Kenji Nakamoto, Jean Philippe Moreau, Priya Kavitha

We present a cross-lingual retrieval model using hierarchical attention and positional alignment for document embedding. It enhances semantic similarity recognition across languages, improving search relevance and retrieval accuracy in multilingual applications with low or inconsistent parallel data availability.

Sarah Olivia Thompson, Michael David Clarke, Emily Rose Sanders, Thomas Gregory Bennett, Robert James Newton

This paper introduces a quantum-assisted attention model for anomaly detection in sensor data. By leveraging entangled feature representations, the system identifies structured anomalies in unlabeled multichannel streams, supporting predictive analytics and fault prevention in real-time industrial monitoring systems.

Liu Zhi Hao, Chen Min Liang, Gao Rui Sheng, Zhang Xiu Lin, Xu Jian Tao, Huang Qiang Wei

This work introduces an interpretable probabilistic framework for completing knowledge graphs. It combines rule embeddings with uncertainty estimates to support robust fact inference and consistency preservation across sparse, incomplete, or noisy knowledge bases.

Thomas Henry Richardson, Rachel Olivia Burke, Sarah Victoria Evans, Benjamin Arthur Campbell, Emily Sophie Palmer