aanaseer / scoehr Goto Github PK

Develop the autoencoder architecture to learn compact representations of input data.

Build the UNet score-network architecture.

Implement the Euler Maruyama method to numerically solve the reverse-time SDE for the sample generation process.

Create data preprocessing pipelines, build datasets, and implement a DataLoader for efficient data loading during training.

Datasets to be implemented:

1. MIMIC-III Clinical Database [1, 2, 3]
2. Yale Emergency Department Dataset [4, 5]

References
[1] Johnson, A., Pollard, T., & Mark, R. (2016). MIMIC-III Clinical Database (version 1.4). PhysioNet. https://doi.org/10.13026/C2XW26.
[2] Johnson, A. E. W., Pollard, T. J., Shen, L., Lehman, L. H., Feng, M., Ghassemi, M., Moody, B., Szolovits, P., Celi, L. A., & Mark, R. G. (2016). MIMIC-III, a freely accessible critical care database. Scientific Data, 3, 160035.(https://www.nature.com/articles/sdata201635)
[3] Goldberger, A., Amaral, L., Glass, L., Hausdorff, J., Ivanov, P. C., Mark, R., ... & Stanley, H. E. (2000). PhysioBank, PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals. Circulation [Online]. 101 (23), pp. e215–e220.
[4] Hong, Woo Suk, Adrian Daniel Haimovich, and R. Andrew Taylor. "Predicting hospital admission at emergency department triage using machine learning." PloS one 13.7 (2018): e0201016.
[5] https://github.com/yaleemmlc/admissionprediction

Implement the forward and reverse-time SDE (Variance-Preserving SDE from [1]) to simulate diffusion processes.

References
[1] Song, Yang, et al. "Score-Based Generative Modeling through Stochastic Differential Equations." International Conference on Learning Representations. 2021.

Create the training loop for the autoencoder model.

Add pre-commit hooks for code linting and formatting to maintain consistent code quality.

Define and implement the de-noising score matching objective.

Develop the training loop for the diffusion model.