aobol / kamnet

KamNet is the state-of-the-art neural network model for spherical liquid scintillator detector.

KamNet

KamNet is the state-of-the-art neural network model for spherical liquid scintillator detector.

Dependencies

All of KamNet's prerequisite comes from the s2cnn packages, we recommend the users to first install that package, and then KamNet will be ready to use. The instruction below are copied from the s2cnn github page:

⚠️ ⚠️ This code is old and does not support the last versions of pytorch! Especially since the change in the fft interface. ⚠️ ⚠️

• PyTorch: http://pytorch.org/ (>= 0.4.0)
• cupy: https://github.com/cupy/cupy
• lie_learn: https://github.com/AMLab-Amsterdam/lie_learn
• pynvrtc: https://github.com/NVIDIA/pynvrtc

(commands to install all the dependencies on a new conda environment)

conda create --name cuda9 python=3.6 
conda activate cuda9

# s2cnn deps
#conda install pytorch torchvision cuda90 -c pytorch # get correct command line at http://pytorch.org/
conda install -c anaconda cupy  
pip install pynvrtc joblib

# lie_learn deps
conda install -c anaconda cython  
conda install -c anaconda requests  

Installation

To install, run

$ python setup.py install

---

Data and Training

• The source code of KamNet is stored in the model folder.
• The data folder contains the pre-processing code to produce the input spatiotemporal data. Pre-processing code is written to be submitted as batch job to the Boston University Shared Computing Cluster system with qsub. Using it on other common batch system (such as slurm) needs additional modifications.
• The spatiotemporal data will be stored in .pickle file as one python dictionary per event, each event contains a list of 2D hit maps following temporal order. Each 2D hit map is stored as CSR sparse matrix to save memory and disk space. Input to KamNet is a .dat list with addresses to all the .pickle files.
• We plan to open source our training data in a stepwise manner, including:
• • Benchmarking dataset: referred to as sim-FAST in the paper, available soon
• • Decay to Excited States dataset: referred to as sim-RAT in the paper, available soon
• • KamLAND-Zen 800 officical MC simulation: referred to as sim-KLZ800 in the paper, need approval from the KamLAND-Zen collaboration

---

Acknowledgement

If you used this model in your work, please cite this paper:

@article{Li:2022frp,
    author = "Li, A. and Fu, Z. and Winslow, L. and Grant, C. and Song, H. and Ozaki, H. and Shimizu, I. and Takeuchi, A.",
    title = "{KamNet: An Integrated Spatiotemporal Deep Neural Network for Rare Event Search in KamLAND-Zen}",
    eprint = "2203.01870",
    archivePrefix = "arXiv",
    primaryClass = "physics.ins-det",
    month = "3",
    year = "2022"
}