Research
Academic publications and peer-reviewed work.
AquaSent-TMMAE: A Self-Supervised Learning Method for Water Quality Monitoring from Spatiotemporal Data
Lee, C., Nabulsi, F., Xu, M., et al.
Abstract
A novel self-supervised learning approach for monitoring water quality using spatiotemporal data. The method leverages masked autoencoders to learn robust representations from environmental sensor data, enabling accurate water quality predictions with limited labeled data.
My Contribution
Contributed to the model architecture design and experimental validation of the temporal masking strategy.
30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining
CLE-SMOTE: Addressing Extreme Imbalanced Data Classification with Contrastive Learning-Enhanced SMOTE
Lee, C., Nabulsi, F., Xu, M., et al.
Abstract
A method combining contrastive learning with SMOTE to handle extreme class imbalance in classification tasks. The approach uses contrastive representations to generate more meaningful synthetic samples, improving classification performance on highly skewed datasets.
My Contribution
Developed the contrastive learning integration and conducted ablation studies on synthetic sample quality.
International Conference on Learning Representations