Autoencoder Optimization for Anomaly Detection: A Comparative Study with Shallow Algorithms

This repository contains the code accompanying the research paper:

"Autoencoder Optimization for Anomaly Detection: A Comparative Study with Shallow Algorithms"

Authors: Vikas Kumar, Vishesh Srivastava, Sadia Mahjabin, Emmanuel Müller

Published in: 2024 International Joint Conference on Neural Networks (IJCNN)

Overview

This project explores the optimization of autoencoders for anomaly detection and compares their performance against traditional shallow algorithms. Specifically, it is compared with the ADBENCH benchmark, and our approach has shown improved accuracy. The code includes implementations of various autoencoder architectures and shallow models, along with scripts to evaluate their effectiveness on benchmark datasets.

Repository Structure

....

Requirements

• Python 3.x
• Required libraries: numpy, pandas, tensorflow, scikit-learn, matplotlib

Install the required libraries using:

pip install -r requirements.txt

Usage

1. Data Preparation: Place your datasets in the data/ directory.
2. Training: Use the scripts in the scripts/ directory to train models.
3. Evaluation: Evaluate the trained models using the provided evaluation scripts.

For detailed instructions, refer to the documentation within each script.

Citation

If you use this code in your research, please cite our paper:

@inproceedings{Kumar2024Autoencoder,
  title={Autoencoder Optimization for Anomaly Detection: A Comparative Study with Shallow Algorithms},
  author={Kumar, Vikas and Srivastava, Vishesh and Mahjabin, Sadia and Müller, Emmanuel},
  booktitle={Proceedings of the International Joint Conference on Neural Networks (IJCNN)},
  year={2024}
}

License

This project is licensed under the MIT License. See the LICENSE file for details.

Contact

For questions or feedback, please contact Vishesh Srivastava at [email protected].