This repository contains the code used in the dissertation Breaking Omnifold by Eliott Menard. It includes:
-
Data Generation Code:
- Generates synthetic datasets as described in the dissertation.
- Features detailed annotations for ease of understanding and reproducibility.
- Note: The T-SNE visualizations are currently incomplete and may not provide meaningful insights due to time constraints.
-
Modified Omnifold Implementation:
- Based on the original Omnifold code from hep-lbdl/OmniFold.
- Incorporates the modifications discussed in the dissertation for re-weighting.
- Includes an additional change aimed at optimizing resource usage for lower-end devices, albeit with some trade-offs in performance.
- Implements an iterative trial framework for training neural networks, as outlined in the dissertation.
- Note: This code lacks thorough documentation, and the implemented changes may not be immediately apparent.
- The data generation code is well-annotated and functional.
- The Omnifold code modifications, while functional, are not properly commented. Additional work may be needed to fully understand the changes.
- The original Omnifold implementation can be found at hep-lbdl/OmniFold.
- For more details on the Omnifold methodology, refer to the official paper: OmniFold: A Method to Simultaneously Unfold All Observables, published in Physical Review Letters.
Feel free to reach out or open an issue if you have questions or suggestions for improvements!