Speaker
Description
In hadron collider experiments, it is essential to accurately reconstruct particles from detector information in order to gain a deeper understanding of a various phenomena. In particular, since detectors such as ATLAS have limited spatial granularity in the calorimeter cell, it is challenging to reconstruct particles within jets.
To mitigate this difficulty, we are developing a Particle Flow algorithm based on an ATLAS-like detector. Specifically, we construct a model that takes the calorimeter cell and track information as input and outputs features of each particle, including its number, particle type, momentum, and direction.
One of the main challenges in this task is that the number of output particles is not known in advance. Conventional approaches often require an additional step to estimate the expected number of particles or to explicitly associate detector objects with truth particles.
Inspired by the DEtection TRansformer (DETR) architecture used for object detection, we propose an end-to-end model that directly generates particle candidates from calorimeter and track. In this poster, we present an overview of our model and its performance on a single-jet benchmark.