We propose a contrastive pretraining approach for semantic segmentation of remote sensing data. The model improves generalization to unlabeled domains by learning invariant features under varied conditions, enhancing segmentation accuracy with limited ground truth in multispectral and hyperspectral datasets.

Chen Wei Liang, Zhang Feng Rui, Xu Tian Ming, Huang Xiu Mei, Gao Lin Zhao, Liu Hong Jie

This paper presents a multilingual sentiment analysis model combining LSTM networks with attention mechanisms. It improves classification in noisy, real-time social media streams by dynamically weighting contextual relevance and capturing cross-linguistic variations across diverse digital expressions.

Thomas Edward Harris, Laura Annette Fisher, Emily Grace Hamilton, David Christopher Bennett, Rachel Olivia Carter

A quantum reinforcement learning system is introduced for controlling autonomous swarms. By optimizing agent behavior using entanglement-like coordination, the model enhances decision accuracy and cooperation efficiency, enabling scalable deployment across dynamic, non-linear, and constraint-heavy environments in real time.

Liu Cheng Wei, Wang Qiang Jun, Zhang Xiang Ning, Xu Min Tao, Huang Bo Rui, Chen Hong Lei

We develop a lightweight GAN framework for generating synthetic samples in rare-event classification tasks. The model stabilizes class distributions, improves minority class recall, and prevents overfitting, offering an efficient augmentation solution for high-risk, low-frequency scenarios.

Emily Charlotte Brooks, Robert Nathaniel King, Olivia Claire Morgan, James Vincent Daniels, Sarah Jane Foster

This paper proposes a dual attention-based representation fusion model for human activity recognition. It aligns visual and sensor data streams, leveraging cross-modal relationships to enhance accuracy in activity classification and enable context-aware automation within smart spaces.

Ravi Subhash Rao, Maria Fernanda Diaz, Pooja Lakshmi Nair, Jean Louis Moreau, Siddharth Raj Deshpande

We present a hierarchical attention fusion model for climate forecasting and anomaly detection. By aggregating features across temporal and spatial scales, the model improves long-range prediction accuracy and facilitates early identification of environmental anomalies using multiresolution climate datasets.

Chen Ming Zhao, Liu Fang Yu, Xu Zhi Rong, Gao Ping Yuan, Zhang Hui Lei

This research introduces an edge-based learning framework that utilizes federated meta-optimization to adapt quickly to local user environments. The approach ensures privacy preservation, efficient model transfer, and high responsiveness for real-time context-aware applications under bandwidth-constrained conditions.

Neha Suresh Menon, Rakesh Gopal Iyer, Taro Kazuki Nakamoto, Carlos Javier Herrera, Elena Sofia Rossi