WlWl Polarization

Measure WWjj polarization fraction

Paper: arXiv:2109.09924
Notice: This code can only be used for the inference process, if you want to train your own model, please contact [email protected].

Requirements

Both Linux and Windows are supported.
64-bit Python3.6(or higher, recommend 3.8) installation.
Tensorflow2.x(recommend 2.6), Numpy(recommend 1.19.5), Matplotlib(recommend 3.4.2)
One or more high-end NVIDIA GPUs(at least 4 GB of DRAM), NVIDIA drivers, CUDA(recommend 11.4) toolkit and cuDNN(recommend 8.2.x).

Preparing dataset

The raw dataset needs to be transformed before it can be imported into the model.

You need to create a raw dataset(we provide a test dataset, stored in ./raw/), the data structure is as follows:

The file has N events:
   Event 1
   Event 2
   ...
   Event N
One event for every 6 lines:
   1. first lepton 
   2. second lepton 
   3. first FB jet 
   4. second FB jet 
   5. MET 
   6. remaining jet 
Each line has the following five columns of elements:
   1.ParticleID  2.Px  3.Py  4.Pz  5.E
The format of an event in the dataset is as follows:
   ...
   -1.0  166.023   5.35817   10.784    166.459
   1.0   -36.1648  -64.1513  -28.9064  79.113
   7.0   -11.3233  -39.6316  -318.178  320.85
   7.0   -34.2795  22.0472   622.79    624.128
   0.0   -22.6711  52.8976   -422.567  426.468
   6.0   -49.9758  29.3283   274.517   294.098
   ...

ParticleID: 1 for electron, 2 for muon, 3 for tau, 4 for b-jet, 5 for normal jet, 0 for met, 6 for remaining jets, 7 for forward backward jet, signs represent electric charge.

Use the command python create_dataset.py YOUR_RAWDATA_PATH, it will create a file with the same name as YOUR_RAWDATA_PATH in the ./dataset/.

Using pre-trained models

After completing the preparation of the dataset, you can use the model to predict the polarization fraction.

Pre-trained weights are placed in ./weights/.
Use the command python inference.py --dataset YOUR_TRADATA_NAME --model_name <MODEL_NAME> --energy_level <ENERGY_LEVEL>, it will give the polarization fractions.

Notice: <ENERGY_LEVEL> should correspond to the collision energy of events.

Example

Run the following command to get the polarization fractions for the standard model:

python create_dataset.py ./raw/sm.dat
python inference.py --dataset sm --model_name TRANS --energy_level 13

Citation

@misc{li2021polarization,
    title={Polarization measurement for the dileptonic channel of $W^+ W^-$ scattering using generative adversarial network},
    author={Jinmian Li and Cong Zhang and Rao Zhang},
    year={2021},
    eprint={2109.09924},
    archivePrefix={arXiv},
    primaryClass={hep-ph}
}