EE340-Project2-DAGMM

Introduction

Repo for EE340 Statistical Learning for Data Science Project 2, SUSTech, 2023 Spring.

For the project requirement, please refer to the requirement. I am sorry that the requirement is only in Chinese.

We choose item 2 (Anomaly Detection 异常检测) and implement the DAGMM model with PyTorch.

DAGMM (Deep Autoencoding Gaussian Mixture Model) is an unsupervised anomaly detection model proposed by Zong et al. in 2018.

"Instead of using decoupled two-stage training and the standard Expectation-Maximization (EM) algorithm, DAGMM jointly optimizes the parameters of the deep autoencoder and the mixture model simultaneously in an end-to-end fashion, leveraging a separate estimation network to facilitate the parameter learning of the mixture model."

Usage

git clone https://github.com/squarezhong/EE340-Project2-DAGMM
cd EE340-Project2-DAGMM
python main.py

It is optional to use pip install -r requirements.txt to install the required packages.

Hyperparameters

The hyperparameters are not fine-tuned (even not tuned at all orz). You can change the hyperparameters in main.py to get better performance.

Reference

Zong, B., Song, Q., Min, M. R., Cheng, W., Lumezanu, C., Cho, D., & Chen, H. (2018). Deep autoencoding Gaussian mixture model for unsupervised anomaly detection. International Conference on Learning Representations (ICLR).