Skip to content

Particle Track Reconstruction with Quantum Algorithms

License

Notifications You must be signed in to change notification settings

cnktysz/HepTrkX-quantum

Repository files navigation

Particle Track Reconstruction with Quantum Algorithms

News:

The final version of this project is released on a different repository. You can access it through this link.

Presented most recent results at Connecting The Dots 2020, talk is here. Proceeding is available as a preprint here

Latest updates are presented at CERN openlab Technical Workshop 2020. Presentation is here.

CHEP 2019 proceeding submitted and available as a preprint here.

How to use?

Use train.py to train a model. Models are available in ./qnetworks folder. Choose the model and other hyperparameters using a configuration file (see ./configs folder for examples).

Execute the following to train the model.

python3 train.py [PATH-TO-CONFIG-FILE]

Note: 1 epoch takes ~1 week when a GPU is used to simulate quantum circuits.

A brief look at the data and the models used

A preprocessed event example

A subgraph example

All events are divided to 8 in $\eta$ direction and 2 in $z$ direction. Therefore, 1 event contains 16 subgraphs. In this work we preprocessed 100 events to create 1600 subgraphs. 1400 is used for training, while 200 is left for validation.

Hybrid Quantum Classical Graph Neural Network Model used:

Quantum Network Structure:

First the data is encoded via a IQC layer, then a PQC is applied to the circuit. Then, the measurements are taken and expectation values are calculated by taking averages of them.

Recent Results

Disclaimer

Notice that this is a project in progress. Latest results are always updated and might be different from the proceeding.

The repository has many scripts which are not complete! Work in Progress!

About

Particle Track Reconstruction with Quantum Algorithms

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published