Machine learning-based anomaly search at the trigger level of a particle detector

OR How do you get rid of 40 million kitchen sponges?

by Sven Bollweg

Imagine this: You're sitting in front of a mountain of 40 million kitchen sponges looking for something. Unfortunately, you don't know what you're looking for, you just know that it looks different from a classic kitchen sponge. In this slam, you will find out how machine learning can help you find the object you are looking for and get rid of the kitchen sponges and why you have 40 million kitchen sponges in the first place.

Curriculum vitae.

Sven is 27 years old and has lived in Hamburg all his life. After graduating from high school, he completed a Bachelor's degree in Computing in Science, specialising in physics, followed by a Master's degree in physics. He is now a doctoral student in the working group for machine learning in particle physics at the University of Hamburg.

On research and work.

Our research includes the development of new algorithms for physics research based on machine learning and artificial intelligence and their application in the search for new physical phenomena in experimental particle physics.


In particle physics, a large number of particle collisions are needed to discover new particles. However, it is not possible to store all collisions. Therefore, a trigger system is needed to filter out the interesting events directly. An anomaly search based on machine learning can be used to search for new particles independently of the model.