DeepZip

Update: Please checkout our new work DZip presented at DCC 2021.

Description

Data compression using neural networks

DeepZip: Lossless Data Compression using Recurrent Neural Networks

Requirements

1. GPU, nvidia-docker (or try alternative installation)
2. python 2/3
3. numpy
4. sklearn
5. keras 2.2.2
6. tensorflow (cpu/gpu) 1.8

(nvidia-docker is currently required to run the code)
A simple way to install and run is to use the docker files provided:

cd docker
make bash BACKEND=tensorflow GPU=0 DATA=/path/to/data/

Alternative Installation

cd DeepZip
python3 -m venv tf
source tf/bin/activate
bash install.sh

Code

To run a compression experiment:

Data Preparation

1. Place all the data to be compressed in data/files_to_be_compressed
2. Run the parser

cd data
./run_parser.sh

Running models

1. All the models are listed in models.py
2. Pick a model, to run compression experiment on all the data files in the data/files_to_be_compressed directory

cd src
./run_experiments.sh biLSTM GPUID

Note: GPUID by default can be set to 0. The corresponding command would be then ./run_experiments.sh biLSTM 0

Please cite if you utilize the code in this repository.

@inproceedings{7fcb664b03ac4d6497048954d756b91f,
title = "DeepZip: Lossless Data Compression Using Recurrent Neural Networks",
author = "Mohit Goyal and Kedar Tatwawadi and Shubham Chandak and Idoia Ochoa",
year = "2019",
month = "5",
day = "10",
doi = "10.1109/DCC.2019.00087",
language = "English (US)",
series = "Data Compression Conference Proceedings",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
editor = "Ali Bilgin and Storer, {James A.} and Marcellin, {Michael W.} and Joan Serra-Sagrista",
booktitle = "Proceedings - DCC 2019",
address = "United States",
}