INR-TPC-Compression

This repository contains the code to reproduce the results presented in the paper: "Efficient Compression of Sparse Accelerator Data Using Implicit Neural Representations and Importance Sampling". The code implements methods to compress sparse accelerator data using various models, including FFNet, SIREN, and WIRE, combined with importance sampling techniques.

Table of Contents

• Installation
• Directory Structure
• Data Description
• Usage
  • Data Preparation
  • Running the Experiments
• Models
• Results
• License
• Citation

Installation

Prerequisites

• Anaconda/Miniconda installed on your system.
• Python 3.8+ (The environment.yml file will handle this)

Setup

1. Clone the repository:

   git clone https://github.com/Xihaier/INR-TPC-Compression
   cd INR-TPC-Compression

2. Create and activate the conda environment:

   conda env create -f environment.yml
   conda activate INR_TPC

3. Verify the installation:

   Ensure that all dependencies are correctly installed by running:

   python -c "import torch; print(torch.__version__)"

Directory Structure

The directory structure of this repository is as follows:

.
├── data/                   # Contains datasets and related files
├── models/                 # Contains model definitions
│   ├── ffnet.py            # FFNet model implementation
│   ├── siren.py            # SIREN model implementation
│   └── wire.py             # WIRE model implementation
├── dataload.py             # Data loading and preprocessing utilities
├── task_1.py               # Script for the first set of experiments
├── task_2.py               # Script for the second set of experiments
├── task_3.py               # Script for the third set of experiments
├── run.sh                  # Shell script to run all tasks sequentially
├── environment.yml         # Conda environment configuration
└── README.md               # Readme file

Data Description

The dataset used in this project contains high-energy physics data from particle tracking detectors, specifically from the Time Projection Chamber (TPC) of the sPHENIX experiment. This dataset records sparse 3D trajectories of charged particles generated in high-energy collisions. Each data point represents a spatial location with varying intensity, capturing the unique patterns of particle paths within the TPC. The data required to run this repository's experiments can be downloaded from Zenodo.

Usage

Data Preparation

Ensure that your data is stored in the data/ directory. You may need to adjust paths in dataload.py or the experiment scripts (task_1.py, task_2.py, task_3.py) depending on your data structure.

Running the Experiments

To reproduce the experiments presented in the paper, simply execute the provided shell script:

bash run.sh

This will sequentially run the experiments as defined in task_1.py, task_2.py, and task_3.py.

Alternatively, you can run each task individually:

python task_1.py
python task_2.py
python task_3.py

Models

The following models are implemented in this repository:

• FFNet: Defined in models/ffnet.py
• SIREN: Defined in models/siren.py
• WIRE: Defined in models/wire.py

These models are designed to compress sparse accelerator data efficiently.

Results

The results of the experiments will be stored in the logs/ directory, which is automatically generated when running the scripts. Each task will create a subdirectory under logs/ containing results, model checkpoints, and plots.

License

This project is licensed under the MIT License.

Citation

Coming soon.