We propose a meta-learning framework for temporal action localization under sparse supervision. Incorporating structured priors and rapid task adaptation, the model achieves accurate event boundary detection in long videos with limited annotations, improving interpretability and generalization across action categories.

Emily Rose Thompson, Nathaniel Gregory Moore, Sarah Amelia Patterson, Robert Jason Ellis, Benjamin Charles Newman

This paper presents a federated attention-based system for predictive modeling in privacy-restricted environments. It allows secure cross-institutional training while retaining local contextual signals, enhancing model generalization for healthcare and insurance applications without compromising user-sensitive data or regional regulations.

Amit Ramesh Kulkarni, Jean Philippe Armand, Priya Sangeetha Iyer, Fatima Laila Mourad, Carlos Antonio Fernandez

We propose a graph-structured time series model using edge-conditioned convolutions and spatial attention. The system captures dependencies between variables in distributed sensor networks, improving accuracy and interpretability in multivariate forecasting tasks across real-world industrial and environmental monitoring scenarios.

Chen Rong Jun, Liu Wen Tao, Xu Ling Fang, Zhang Yue Liang, Gao Ming Rui, Huang Bo Lin

This research introduces a contrastive adaptation approach for multi-domain learning in NLP. By aligning latent spaces across languages, the model reduces domain gaps and improves transfer efficiency for downstream tasks including classification, summarization, and entity recognition under multilingual constraints.

Emily Rose Hamilton, Thomas Arthur Simmons, Olivia Jade Clarke, Rachel Anne Whitmore, James Robert Foster

This paper proposes a reinforcement learning model enhanced with domain knowledge to optimize resource allocation across edge-cloud layers. The system ensures task prioritization, latency control, and adaptive scaling, improving operational efficiency under dynamic workloads in heterogeneous computing environments.

Neha Vidhya Raghavan, Jean Louis Dufort, Taro Kenta Ishikawa, Siddharth Arvind Rao, Maria Paula Gutierrez

This study presents a quantum-enhanced metric learning model that improves class separation in low-dimensional embeddings. By leveraging quantum kernels, the model enhances contrastive learning, enabling efficient and discriminative representation generation for visual classification tasks on compact datasets.

Liu Xiang Wen, Zhang Hao Ming, Gao Rui Sheng, Chen Ming Zhe, Xu Qiang Liang, Huang Jie Bo

This study presents an interpretable deep learning model for generating radiology reports. The architecture leverages hierarchical co-attention and structured concept mapping to produce clinically relevant narratives from image findings, ensuring trust and transparency in automated diagnostics.

Liu Yi Chang, Xu Heng Jie, Gao Xiang Rui, Chen Min Liang, Zhang Yue Xin, Huang Li Bo