Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Veli: Unsupervised Method and Unified Benchmark for Low-Cost Air Quality Sensor Correction
Yahia Dalbah, Marcel Worring, Yen-Chia Hsu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| apply Veli instead of raw LCS readings | Low-cost sensors · In-distribution Utrecht | MAE: 24.77 -> 5.25 | Improved |
| apply Veli instead of raw LCS readings | Low-cost sensors · In-distribution Rotterdam | MAE: 21.27 -> 3.36 | Improved |
| apply Veli to already accurate sensors | Low-cost sensors · In-distribution Nijmegen | MAE: 2.82 -> 3.06 | Hurt |
| use Veli instead of Kalman filter denoising | KF · In-distribution Utrecht | MAE: 15.95 -> 5.25 | Improved |
| fine-tune the encoder on target-region data with decoder frozen (30 epochs) | Veli · Out-of-distribution Tainan | MAE: 8.59 -> 7.83 | Improved |
| fine-tune the encoder on target-region data with decoder frozen (30 epochs) | Veli · Out-of-distribution Taichung | MAE: 7.78 -> 7.65 | Improved |
| apply zero-shot Veli instead of raw LCS | Low-cost sensors · Out-of-distribution Tainan | MAE: 14.09 -> 8.59 | Improved |
| apply KF denoising instead of raw LCS | Low-cost sensors · Out-of-distribution Taichung | MAE: 10.01 -> 9.98 | No clear effect |
| inject NA readings into 1 to 9 of the 10 sensors (ablation) | Veli · In-distribution test sites | MAE | Hurt |
| reduce number of sensors from 10 to 3, 5 or 7 (ablation) | Veli · In-distribution test sites | MAE | No clear effect |
| set beta_y to 0 (ablation) | Veli · Loss weight sensitivity | training loss | Hurt |
| lower beta_z relative to other loss weights (ablation) | Veli · Loss weight sensitivity | MAE | Hurt |
Closest papers by meaning
- AirRadar: Inferring Nationwide Air Quality in China with Deep Neural Networks (AAAI 2025 · similarity 0.82)
- REOBench: Benchmarking Robustness of Earth Observation Foundation Models (NeurIPS 2025 · similarity 0.80)
- AirPhyNet: Harnessing Physics-Guided Neural Networks for Air Quality Prediction (ICLR 2024 · similarity 0.80)
- Air Quality Prediction with Physics-Guided Dual Neural ODEs in Open Systems (ICLR 2025 · similarity 0.80)
- RadarQA: Multi-modal Quality Analysis of Weather Radar Forecasts (NeurIPS 2025 · similarity 0.79)
- UrbanDataLayer: A Unified Data Pipeline for Urban Science (NeurIPS 2024 · similarity 0.79)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
