Publications

Machine Learning Track

• Tree-Regularized Tabular Embeddings, NeurIPS Workshop, 2023. paper, poster, code

• Meta and Contrastive Learning Improves Tabular AUC with Unlabeled Data, Amazon Machine Learning Conference (AMLC), 2022

  • justified the effectiveness and generalizability of semi-supervised and self-supervised framework on two tabular application in Amazon. The proposed methods outperform production models.
  • unprecedentedly bridged the gap between maximizing AUC and minimizing contrastive loss (CL) with empirical and theoretical proofs. Intuitively, it explains that CL can be regarded as surrogate when targetting towards AUC maximization.

Engineering Track

• Theoretical Analysis on Strain Transfer Error of FBG Sensors Attached on Steel Structures Subjected to Fatigue Load. link

• A wireless fatigue monitoring system utilizing a bio-inspired tree ring data tracking technique. link

• Towards an Occupancy-Enhanced Building HVAC Control Strategy Using Wi-Fi Probe Request Information. link
• An optimized FBG-based fatigue monitoring strategy on deepwater risers. link
• Optimization analysis and experimental validation of distributed optical fiber sensors for pavement monitoring based on strain transfer mechanism. link