praeclarumjj3 / ATLAS_Evaluation

Submission for the Evaluation Exercise for ATLAS Encoders

ATLAS Evaluation

[Presentation]

This repo contains the code for my submission of the task given by mentors for the GSoC 2021 Project: Deep autoencoders for ATLAS data compression under CERN-HSF Organization.

Problem Statement: Prepare an autoencoder to compress the four-momentum of a sample of simulated ‘j’ particles from 4 to 3 variables for the dataset available on drive.

Contents

1. Setup Instructions
2. Repository Overview
3. Results

1. Setup Instructions

• Clone the repo:

git clone https://github.com/praeclarumjj3/ATLAS_Evaluation.git

• To run the AE Compression model and see the results, open AE_Compression_3D.ipynb and run it cell by cell.

2. Repository Overview

The repository is structured as follows:

• data - Contains the datas files for training/testing the AE Compression model.
• plotInput - Contains images of the plots for normalized Input Data.
• plotOutput - Contains images of the plots for normalized Input Data.
• prepare_datset.py - Script for cleaning and preparing the .pkl file which contains the data used for training.
• AE_Compression_3D.ipynb - Contains code for running the model.

3. Results

• The compression model works well on the given dataset as can be seen from the overlapping output-input plots for the normalized data: