This paper explores the development of self-evolving AI agents capable of learning goals and adapting dynamically using memory-augmented neural architectures. It supports strategic planning and learning in digital ecosystems with evolving objectives and unknown environmental parameters.

William Alexander Hayes, Emily Rose Preston, Nathan George Thompson, Laura Abigail Jensen, Benjamin Isaac Clarke

We propose a hybrid learning system combining multi-agent reinforcement strategies with federated privacy-preserving protocols. The architecture facilitates decentralized optimization, robust policy convergence, and secure real-time collaboration in intelligent networks, addressing scalability and trust issues in dynamic distributed environments.

Amit Suresh Nair, Jean Claude Laurent, Taro Masaki Nishida, Priya Meena Reddy, Maria Gabriella Santiago

This study presents a time-conditioned attention framework to address irregular sequential patterns in personalized health data. By learning temporal dependencies and adjusting for gaps, the model enhances prediction accuracy and supports continuous monitoring in sensitive, event-driven applications like early risk detection.

Sarah Elizabeth Moore, Benjamin Charles Foster, Olivia Jane Simmons, Robert William Greyson, Laura Isabelle Peterson

This research introduces a dual contrastive learning approach for crossmodal action recognition. By aligning audio and visual cues through self-supervised embedding strategies, the model improves multimodal understanding and enables generalization to novel scenarios with limited labeled data availability.

Chen Wei Qiang, Liu Jian Ping, Gao Xun Feng, Zhang Lin Rui, Xu Bo Han

This paper proposes efficient knowledge distillation techniques for compressing deep models used in industrial automation. It enables high-performance edge deployment without compromising model accuracy, ensuring rapid inference and reduced resource consumption for intelligent control applications across manufacturing infrastructures.

David Anthony Brooks, Emily Rose Carter, Rachel Olivia Greene, Nathan William Clarke, Thomas Gregory Wright

This study proposes a graph neural architecture to detect fraudulent patterns in financial networks. It employs relational aggregation and semi-supervised learning to perform well with limited labels, offering a scalable and interpretable solution for complex anomaly detection in finance.

Zhang Xiao Lei, Liu Feng Xin, Huang Wen Zhi, Xu Chao Ling, Gao Peng Rui, Yang Jie Ming

This work introduces a transformer-based semantic segmentation model tailored for cross-domain applications. It leverages hierarchical attention to enhance spatial coherence and reduce noise sensitivity across multisource imagery, improving adaptability and label efficiency in dynamic visual domains with varying imaging conditions.

Wei Zhang Hao, Xu Ping Liang, Liu Ming Jie, Chen Fang Zhou, Huang Rui Wen, Gao Lin Cheng