Adaptive Graph Convolutional Recurrent Network for Traffic Forecasting

Authors: LEI BAI, Lina Yao, Can Li, Xianzhi Wang, Can Wang

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments on two real-world traffic datasets show AGCRN outperforms state-of-the-art by a significant margin without pre-defined graphs about spatial connections.
Researcher Affiliation Academia Lei Bai UNSW, Sydney baisanshi@gmail.com Lina Yao UNSW, Sydney lina.yao@unsw.edu.au Can Li UNSW, Sydney can.li4@student.unsw.edu.au Xianzhi Wang University of Technology Sydney xianzhi.wang@uts.edu.au Can Wang Griffith University can.wang@griffith.edu.au
Pseudocode No No pseudocode or algorithm blocks were found.
Open Source Code Yes Code available at: https://github.com/Lei BAI/AGCRN
Open Datasets Yes To evaluate the performance of our work, we conduct experiments on two public real-world traffic datasets: Pe MSD4 and Pe MSD8 [6, 11]. Pe MS means Caltrans Performance Measure System (Pe MS) [38]
Dataset Splits Yes We split the datasets into training sets, validation sets, and test sets according to the chronological order. The split ratio is 6:2:2 for both datasets.
Hardware Specification Yes All the deep-learning-based models, including our AGCRN, are implemented in Python with Pytorch 1.3.1 and executed on a server with one NVIDIA Titan X GPU card.
Software Dependencies Yes All the deep-learning-based models, including our AGCRN, are implemented in Python with Pytorch 1.3.1 and executed on a server with one NVIDIA Titan X GPU card.
Experiment Setup Yes We optimize all the models by Adam optimizer for a maximum of 100 epochs and use an early stop strategy with the patience of 15. The best parameters for all deep learning models are chosen through a carefully parameter-tuning process on the validation set.