A novel quantum kernel expansion method is presented to address pattern recognition in non-linear high-dimensional spaces. The system improves generalization, reduces overfitting, and ensures computational efficiency when classifying complex datasets, outperforming traditional support vector approaches in benchmark tasks.

Michael James Wilson, Robert Anthony King, Anna Elizabeth Jensen, Thomas Richard Bennett, Laura Michelle Evans

This study presents a multiscale autoencoder pipeline for semantic segmentation in satellite image streams. The system reduces data redundancy and improves boundary localization accuracy, enhancing spatiotemporal analysis for continuous Earth observation applications under bandwidth and storage-constrained deployments.

Liu Zhen Hui, Yang Bo Lin, Zhang Xiao Rui, Wang Qing Mei, Li Jia Cheng

A deep Bayesian network framework is proposed for forecasting in real-time monitoring systems. It quantifies predictive uncertainty and enables early detection of anomalies in volatile environments, supporting critical decision-making through dynamic threshold adjustment and multi-step ahead estimation.

Rachel Elizabeth Dawson, Jackson Leo Whitman, Benjamin Edward Carter, Olivia Lauren Adams

This paper presents a dual privacy-preserving pipeline integrating differential privacy with homomorphic encryption. The system ensures secure learning on sensitive data without compromising model accuracy, offering a practical solution for privacy-focused industries and decentralized collaborative intelligence.

Amit Rajendra Kulkarni, Jean-Paul Henri Moreau, Divya Shankar Menon, Maria Claudia Navarro

We introduce a hierarchical attention mechanism tailored to time-series forecasting. It models short-term and long-term dependencies with interpretable attention layers, achieving high accuracy in financial and operational sequences while offering insight into temporal feature importance and prediction drivers.

Emily Catherine Brooks, Ryan Nicholas Summers, Laura Isabelle Peterson, Thomas Benjamin Grant

This paper proposes a low-latency sensor fusion model utilizing attention-based graph convolutions. The model optimizes distributed inference by preserving spatial coherence and reducing communication overhead in edge-to-cloud networks for real-time decision support across constrained environments.

Wang Li Sheng, Gao Yi Feng, Zhang Jun Tao, Xu Wen Qiang, Liu Ming Zhe