This paper presents a self-supervised learning framework that captures temporal dependencies from sparse sensor streams. It utilizes contrastive encoding and sequential consistency loss, enabling high-fidelity reconstruction of temporal patterns in data-constrained environments without reliance on external labels or ground truth.

Amit Rajeev Mehta, Anna Sophia Müller, Siddharth Harish Desai, Taro Kenji Nakamura, Maria Julia Navarro, Carlos Felipe Ruiz

This research explores cross-domain transfer in graph neural networks using subspace alignment and structural regularization. The system efficiently adapts representations between domains with minimal labeled data, achieving high performance in heterogeneous graph classification and dynamic topology prediction tasks.

Chen Hao Ming, Li Xin Wei, Zhang Qiang Rui, Liu Mei Zhen, Gao Rui Yuan, Yang Ping Bo

This paper introduces a reinforcement learning method that enhances policy robustness by distilling stochastic actions with contrastive regularization. The model stabilizes learning in environments with noise and volatility, delivering consistent decision outcomes under uncertain and partially observable conditions.

Olivia Jane Mitchell, Thomas Gerald Adams, James Patrick Clark, Michael Brian Foster, Emily Susan Peterson

We propose a hybrid forecasting framework combining variational autoencoders and transformers for anomaly detection. It handles missing data, fuses multiple sensing channels, and delivers high sensitivity in dynamic forecasting tasks under complex environmental fluctuations and limited supervision.

Neha Supriya Kulkarni, Jean Paul François Morel, Ravi Siddharthan Nair, Carlos Enrique Soto, Elena Beatriz Rossi

We propose a dual-attention framework for improving caption generation in noisy multimodal environments. The system aligns textual and visual features using contextual gating and temporal filtering, delivering robust outputs in scenarios with high interference and partial occlusion.

Sarah Amelia Dawson, Robert Francis Wright, Laura Katherine Simmons, David Matthew Cooper, Emma Claire Bennett