/ICFA-Beam-2022

This repository contains all materials from tutorial to anomaly detection and our presentation at 3rd ICFA Beam Dynamics Mini-Workshop on Machine Learning Applications for Particle Accelerators

Primary LanguageJupyter Notebook

Antonin Sulc, website : sulcantonin.github.io

https://indico.bnl.gov/event/16158/

This is a website which was put on for purpose to share our materials for the tutorial Tutorial on Anomaly Detection at 3rd ICFA Beam Dynamics Mini-Workshop on Machine Learning Applications for Particle Accelerators (https://www.bnl.gov/mlaworkshop2022/)

Tutorial: Anomaly Detection

(pdf tutorial slides)

(ipynb basics) (ipynb autoencoder) (ipynb one class loss) (ipynb bce) (ipynb semi-supervised anomaly loss) (ipynb anogan)

If you do not want to install Jupyter or run e.g. Google Colab, you can simply execute the notebook via (https://mybinder.org/).

Or you can use following link which was prepared for us (thank you very much Jon E.) https://www.sirepo.com/jupyterhublogin

Talk: Status of Data-Driven Beam Trajectory Anomaly Detection at European XFEL

(pdf talk slides)

Code + Data will follow soon!

IPAC'22 Conference Contribution https://github.com/sulcantonin/BPM_IPAC22

An accurate assessment of beam orbits is a reliable characteristic of operations at SASE beamlines at European XFEL. The current availability of hardware for data-driven models like GPUs enabled to deploy models which are capable of handling enormous throughput of data provide tools for more sophisticated analysis of current operations with state-of-the-art machine learning models. In this work, we are showing our current progress of beam orbit analysis at European XFEL with pure model-free data-driven tools. We examine more advanced models which take the intra-bunching properties into consideration and analyze beam operations with various anomaly detection methods.