🚩 For example on a small dataset that can run on a personal laptop in example_test. The ./example_test/dataset/ can download from dataset.
Machine Learning-Enhanced High-Resolution Exposure Assessment of Ultrafine Particles
finalcode
├─ CAMS_downscale
│ ├─ CAMSpollutionDistribution.m
│ ├─ CAMS_downscale_plot.ipynb
│ ├─ cams_train.py
│ ├─ cams_train_pred.ipynb
│ ├─ downScaleConc.m
│ ├─ preprocessing.py
├─
B439
PNC_estimate
│ ├─ diffplot.m
│ ├─ PNC_estimate_plot.ipynb
│ └─ swissPNCDistribution.m
├─ PNC_pred
│ ├─ model_train.ipynb
│ ├─ pnc_pred.py
│ ├─ PNC_pred_plot.ipynb
│ ├─ test_pnc_model
│ │ ├─ model_test.py
│ └─ trained model
├─ pop_PNC
│ ├─ pop_data_processing.py
│ ├─ pop_plot_final.ipynb
└─ SI_figure
- downScaleConc.m : If predictionFlag = 0 can get the pollution_trainData.csv, and if predictionFlag = 1 get 8760 hourId_pollutant_predData.bin files.
- preprocessing.py : Contains functions cams_train.py may used.
- cams_train.py : Use pollution_trainData.csv and hourId_pollutant_predData.bin files to the annual pollution high-resolution 1km grid data used machine learning methods.
- cams_train_pred.ipynb: Get site-specific prediction results.
- CAMS_downscale_plot.ipynb : Figure plot code for pollution compares.
- CAMSpollutionDistribution.m : Figure plot code for DownScales pollution.
- swissPNCDistribution.m : The code to plot 2020_DownScale_PNC.png.
- PNC_estimate_plot.ipynb : The code to plot other figures.
- diffplot.m : The code plot diff between matflies.
- test_pnc_model/model_test.py : DL code for pnc_pred.py.
- pnc_pred.py : Get the pnc_estimate mat files.
- model_train.ipynb : Train the PNC models.
- pop_data_processing.py: The functions pop_plot_final.ipynb used.
- pop_plot_final.ipynb: Get the population related figures.
- CPU： Intel(R) Core(TM) i7-10750H
- GPU： NVIDIA Geforce GTX 1650 Ti
- RAM： 16GB
- CPU： Intel(R) Xeon(R) Platinum 8383C
- GPU： NVIDIA Geforce GTX 3090
- RAM： 256GB
- CPU： Intel(R) Xeon(R) Platinum 8488C 2.40 GHz
- GPU： 4 x NVIDIA RTX A6000
- RAM： 512GB
- Operating System： Windows 11
- Python Version： Anaconda Python 3.9.7
- CUDA Version： 11.7
- Operating System： Windows 10
- Python Version： Anaconda Python 3.11.4
- CUDA Version： 12.2
- Operating System： Windows 10
- Python Version： Python 3.12.6
- CUDA Version： 12.6
- For convenience, execute the following command.
pip install -r requirements.txt
-
Start the dataset. A small dataset
./example_test/dataset/can download from dataset as an example. -
Use the
./PNC_pred/model_train.ipynbto train your own model.
If you find this repo useful, please cite our paper.
- [1] Jianyao, Y., Yuan, H., Su, G. et al. Machine learning-enhanced high-resolution exposure assessment of ultrafine particles. Nat Commun 16, 1209 (2025). https://doi.org/10.1038/s41467-025-56581-8.
- [2] Jianyao, Y. & Zhang, X. Machine learning-enhanced high-resolution exposure assessment of ultra fine particles. Zenodo https://doi.org/10.5281/zenodo.14554168 (2024).
-
- [3] Jianyao, Y. & Zhang, X. Machine learning-enhanced high-resolution exposure assessment of ultra fine particles (Mini Dataset + Stem-PNC weights). Zenodo https://doi.org/10.5281/zenodo.14634738 (2025).
If you have any questions or suggestions, feel free to contact:
- Yudie Jianyao (Ph.D. student, jyyd23@mails.tsinghua.edu.cn)