FourierGNN: Rethinking Multivariate Time Series Forecasting from a Pure Graph Perspective
Authors: Kun Yi, Qi Zhang, Wei Fan, Hui He, Liang Hu, Pengyang Wang, Ning An, Longbing Cao, Zhendong Niu
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on seven datasets have demonstrated our superior performance with higher efficiency and fewer parameters compared with state-of-the-art methods. |
| Researcher Affiliation | Academia | 1Beijing Institute of Technology, 2Tongji University, 3University of Oxford 4University of Macau, 5He Fei University of Technology, 6Macquarie University |
| Pseudocode | No | The paper describes methods through mathematical equations and architectural diagrams but does not include structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Code is available at this repository: https://github.com/aikunyi/Fourier GNN. |
| Open Datasets | Yes | We use seven public multivariate benchmarks for multivariate time series forecasting and these benchmark datasets are summarized in Table 7. (Table 7 lists datasets like ECG, Solar, Electricity, COVID-19, METR-LA with URLs/citations in footnotes) and ECG3: This dataset is about Electrocardiogram(ECG) from the UCR time-series classification archive [38]. |
| Dataset Splits | Yes | Except the COVID-19 dataset, we split the other datasets into training, validation, and test sets with the ratio of 7:2:1 in a chronological order. For the COVID-19 dataset, the ratio is 6:2:2. |
| Hardware Specification | Yes | All experiments are conducted in Python using Pytorch 1.8 [37] (except for SFM [24] which uses Keras) and performed on single NVIDIA RTX 3080 10G GPU. |
| Software Dependencies | Yes | All experiments are conducted in Python using Pytorch 1.8 [37] (except for SFM [24] which uses Keras) |
| Experiment Setup | Yes | Our model is trained using RMSProp with a learning rate of 10 5 and MSE (Mean Squared Error) as the loss function. ... Specifically, the embedding size and batch size are tuned over {32, 64, 128, 256, 512} and {2, 4, 8, 16, 32, 64, 128} respectively. |