Robust Spatiotemporal Traffic Forecasting with Reinforced Dynamic Adversarial Training
Machine learning-based forecasting models are commonly used in Intelligent Transportation Systems (ITS) to predict traffic patterns and provide city-wide services. However, these models are susceptible to adversarial attacks, which can lead to inaccurate predictions and negative consequences such as congestion and delays. Improving the adversarial robustness of these models is crucial in ITS. In this paper, we propose a novel framework for incorporating adversarial training into spatiotemporal traffic forecasting tasks. We demonstrate that traditional adversarial training methods in static domains cannot be directly applied to traffic forecasting tasks, as they fail to effectively defend against dynamic adversarial attacks. We propose a reinforcement learning-based method to learn the optimal node selection strategy for adversarial examples, which improves defense against dynamic attacks and reduces overfitting. Additionally, we introduce a self-knowledge distillation regularization to overcome the "forgetting issue" caused by constantly changing adversarial nodes during training.
We evaluate our approach on two real-world traffic datasets and demonstrate its superiority over other baselines. Our method effectively enhances the adversarial robustness of spatiotemporal traffic forecasting models.
This repository includes:
- Code for the RDAT in our study.
Introduction
Environment
Datasets
We use the PeMS-Bay and PeMS-D4 datasets (link).
Usage
Adversarial Robustoness
To train and evaluate a baseline model, run the following commands:
# PeMS-BAY
# pre-train policy net
python train.py configs/PeMS/PeMS-train0.7-val0.1-test0.2-in12out12-STpgd0.1nodes0.5eps-random-none_dropout-gwnet-AT_policy_atten_epoch100-steps10-constant1e-03-start_index_learn-baseline_saliency.yaml -d 0 #device id
# AT
python train.py configs/PeMS/PeMS-train0.7-val0.1-test0.2-in12out12-STpgd0.1nodes0.5eps-random-none_dropout-gwnet-AT_policy_atten_epoch100-steps10-constant1e-03-start_index_learn-baseline_saliency-offine-reg0.4-Exps.yaml -d 0 #device id
# PeMS-D4
# pre-train policy net
python train.py configs/PeMSD4/PeMSD4-train0.7-val0.1-test0.2-in12out12-STpgd0.1nodes0.5eps-none_dropout-gwnet-AT_policy_atten_epoch100-steps30-constant1e-03-start_index_learn-baseline_saliency-offine.yaml -d 0 #device id
# AT
python train.py configs/PeMSD4/PeMSD4-train0.7-val0.1-test0.2-in12out12-STpgd0.1nodes0.5eps-random-none_dropout-gwnet-AT_policy_atten_epoch100-steps30-constant1e-03-start_index_learn-baseline_saliency-offine-dist0.4.yaml -d 0 #device id
References
If you find the code useful for your research, please consider citing
@inproceedings{fan2023RBDAT,
author = {Liu, Fan and Zhang, Weijia and Liu, Hao},
title = {Robust Spatiotemporal Traffic Forecasting with Reinforced Dynamic Adversarial Training},
year = {2023},
isbn = {9798400701030},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3580305.3599492},
doi = {10.1145/3580305.3599492},
booktitle = {Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining},
pages = {1417–1428},
numpages = {12},
keywords = {adversarial learning, adversarial training, robust spatiotemporal traffic forecasting},
location = {Long Beach, CA, USA},
series = {KDD '23}
}and/or our related works
@inproceedings{fan2022ASTFA,
author = {Fan LIU, Hao LIU, Wenzhao JIANG},
title = {Practical Adversarial Attacks on Spatiotemporal
Traffic Forecasting Models},
booktitle = {In Proceedings of the Thirty-sixth Annual Conference on Neural Information Processing Systems (NeurIPS)},
year = {2022}
}Acknowledgement
We thank the authors for the following repositories for code reference: Adversarial Long-Tail, etc.