A research project that uses Normalizing Flows to perform Background Estimation Task. (This project was carried out for GSoC-2021 (Google Summer of Code) with ML4SCI (Machine Learning for Science), Click Here for more details.).
Explore the docs »
·
Report Bug
·
Request Feature
Data-driven background estimation is crucial for many scientific searches, including searches for new phenomena in experimental datasets. Neural autoregressive flows (NAF) is a deep generative model that can be used for general transformations and is therefore attractive for this application. The MLBENDER project focuses on studying how to develop such transformations that can be learned and applied to a region of interest. In this project the main aim is implementing a Neural Autoregressive Flow (NAF) model to estimate the background distribution and apply it to a representative physics analysis searching for a resonance excess over a smooth background.
- All project related informations can be found in main.ipynb file.
It is recommended to use conda or venv environments if you're going to run this on your local PC.
Main packages are as follows:
- PyTorch 1.9
- Python 3.6 or higher
- Pandas
- Seaborn
- numpy
- sklearn
- Dataset can be found at: LHC Dataset.
- You can work with the other datasets if you like, the normalizing flow models are suitable to work with any other datasets.
See the open issues for a list of proposed features (and known issues).
Contributions are what make the open source community such an amazing place to be learn, inspire, and create. Any contributions you make are greatly appreciated.
- Fork the Project
- Create your Feature Branch (
git checkout -b feature/AmazingFeature) - Commit your Changes (
git commit -m 'Add some AmazingFeature') - Push to the Branch (
git push origin feature/AmazingFeature) - Open a Pull Request
Distributed under the MIT License. See LICENSE for more information.
Sinan Gençoğlu - @SinanGncgl - [email protected]
Project Link: https://github.com/SinanGncgl/Background-Estimation-NormalizingFlows