We develop an optimized quantum kernel-based classifier for identifying complex non-linear class boundaries. The approach leverages quantum feature space expansion and iterative projection optimization to achieve high accuracy and robustness in challenging multi-class classification tasks across synthetic and real datasets.

Zhang Hao Wei, Li Fang Yuan, Xu Hui Rong, Gao Xiu Lin, Yang Ming Cheng

A latent variable inference approach is proposed to facilitate cross-domain generalization in medical image classification. The model improves transfer performance between datasets with minimal labels, providing a scalable solution for domain adaptation under distributional shifts.

Emily Susan Walker, James Edward Preston, Sarah Amelia Dawson, Robert Patrick Grant, Olivia Alice Middleton

We introduce a real-time visual intelligence system that fuses multiscale thermal and hyperspectral data for enhanced scene analysis. The framework adapts dynamically to streaming conditions, improving resolution and classification fidelity in environments with complex optical signatures.

Liang Cheng Long, Wang Tian Bo, Zhang Wei Ming, Liu Rong Zhao, Huang Fei Sheng

This study presents a federated adaptive learning approach to securely train models across institutions without centralizing data. It enables personalized performance, reduces communication overhead, and supports multi-institutional collaboration through adaptive weight updates and encrypted aggregation mechanisms.

Ramesh Pranav Iyer, Fatima Noor Al-Mansoori, Jean Claude Bernard, Pooja Meenakshi Sharma, Elena Cristina Rossi, Ivan Viktor Petrov

This work compares vision transformers and graph-based reasoning models in visual relationship analysis tasks. Focusing on interpretability, the models are benchmarked on structured datasets, showing significant differences in attention behavior, inference speed, and classification granularity across domain-specific contexts.

James Alexander Collins, Emily Theresa Walker, Rachel Isabelle Brooks, Thomas William Greene, Benjamin Lucas Foster

We introduce quantum-inspired attention mechanisms to model spatiotemporal dynamics in surveillance environments. The system integrates heterogeneous sensor streams and dynamically highlights key frames using adaptive attention heads, enhancing recognition accuracy under variable signal quality conditions and shifting perspectives.

Liu Zhi Hao, Zhang Ming Jie, Huang Wei Cheng, Wang Bo Qiang, Xu Lei Yun, Gao Xiang Ling

This paper presents a temporal graph embedding framework for predictive maintenance in multi-agent networks. Probabilistic modeling enables early fault detection and reduces false positives, enhancing system reliability through dynamic relational learning and structural awareness across decentralized industrial environments.

Aditya Pranav Joshi, Carlos Antonio Navarro, Nikhil Shankar Iyer, Fatima Sofia Al-Muhammad, Elena Cristina Rossi