Neural Transformation Learning for Anomaly Detection (NeuTraLAD)

This is the official code for a PyTorch implementation of Neural Transformation Learning reported in the paper Neural Transformation Learning for Deep Anomaly Detection Beyond Images by Chen Qiu et al. The paper is published in ICML 2021 and can be found here https://arxiv.org/abs/2103.16440. The code allows the users to reproduce and extend the results reported in the study. Please cite the above paper when reporting, reproducing or extending the results.

Purpose of the project

This software is a research prototype, solely developed for and published as part of the publication cited above. It will neither be maintained nor monitored in any way.

Reproduce the Results

This repo contains the code of experiments with NeuTraLAD on various data types including time series data (one-vs-rest setting), tabular data, image data (one-vs-rest setting).

Please run the command and replace $# with available options (see below):

python Launch_Exps.py --config-file $1 --dataset-name $2 

config-file:

• Tabular Data: config_thyroid.yml; config_arrhy.yml; config_kdd.yml; config_kddrev.yml;

• Time Series: config_arabic.yml; config_characters.yml; config_natops.yml; config_epilepsy.yml; config_rs.yml;

• Image Data: config_fmnist.yml; config_cifar10_feat.yml;

dataset-name:

• Tabular Data: thyroid; arrhythmia; kdd; kddrev;

• Time Series: arabic_digits; characters; natops; epilepsy; racket_sports;

• Image Data: fmnist; cifar10_feat;

How to Use

1. When using your own data, please put your data files under DATA.

2. Create a config file which contains your hyper-parameters under config_files.

3. Add your data loader to the function ''load_data'' in the loader/LoadData.py.

• For time series data, the shape is (batch size, #channels, sequence length).
• For image data, the shape is (batch size, #channels, height, width).
• For tabular data/features from pre-trained model, the shape is (batch size, feature dim).

Datasets

• Time series datasets are downloaded from the UEA datasets from https://www.timeseriesclassification.com/. The processed data of NATOPS, Epilepsy, and Racket_Sports are available under DATA.

• Arrhythmia and Thyroid datasets are downloaded from https://github.com/lironber/GOAD. Please put the data under DATA.

• KDD and KDDrev datasets are downloaded from https://kdd.ics.uci.edu/databases/kddcup99/. Please put the data under DATA.

• Cifar10_feat is the last-layer features of Cifar 10 extracted by a ResNet152 pretrained on ImageNet. Extract_img_features.py is used to extract features.

Citation

If you use this work please cite

@inproceedings{qiu2021neural,
  title={Neural transformation learning for deep anomaly detection beyond images},
  author={Qiu, Chen and Pfrommer, Timo and Kloft, Marius and Mandt, Stephan and Rudolph, Maja},
  booktitle={International Conference on Machine Learning},
  pages={8703--8714},
  year={2021},
  organization={PMLR}
}

License

Neural Transformation Learning for Anomaly Detection (NeuTraLAD) is open-sourced under the AGPL-3.0 license. See the LICENSE file for details.

For a list of other open source components included in Neural Transformation Learning for Anomaly Detection (NeuTraLAD), see the file 3rd-party-licenses.txt.