SCINet: Time Series Modeling and Forecasting with Sample Convolution and Interaction

Authors: Minhao LIU, Ailing Zeng, Muxi Chen, Zhijian Xu, Qiuxia LAI, Lingna Ma, Qiang Xu

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results show that SCINet achieves significant forecasting accuracy improvements over both existing convolutional models and Transformer-based solutions across various real-world time series forecasting datasets.
Researcher Affiliation Academia Minhao Liu , Ailing Zeng, Muxi Chen, Zhijian Xu, Qiuxia Lai, Lingna Ma, Qiang Xu CUhk REliable Computing (CURE) Lab. Dept. of Computer Science & Egnineering, The Chinese University of Hong Kong {mhliu,qxu}@cse.cuhk.edu.hk
Pseudocode No The paper describes the model architecture and processes with figures and equations, but it does not include a dedicated pseudocode or algorithm block.
Open Source Code Yes Our codes and data are available at https://github.com/cure-lab/SCINet.
Open Datasets Yes We conduct experiments on 11 popular time series datasets: (1) Electricity Transformer Temperature [42] (ETTh) (2) Traffic (3) Solar-Energy (4) Electricity (5) Exchange-Rate (6) Pe MS (PEMS03, PEMS04, PEMS07 and PEMS08). A brief description of these datasets is listed in Table 1. All the experiments on these datasets in this section are conducted under multi-variate TSF setting. To make a fair comparison, we follow existing experimental settings, and use the same evaluation metrics as the original publications [17, 26, 40, 19] in each dataset. ... Data partition Follow [42] Training/Validation/Testing: 6/2/2
Dataset Splits Yes Data partition Follow [42] Training/Validation/Testing: 6/2/2
Hardware Specification No The paper states in its checklist that hardware specifications are included in the Appendix ('See the Appendix'), but the Appendix content is not provided for analysis in this excerpt. The main body of the paper does not specify any particular hardware (e.g., specific GPU or CPU models) used for experiments.
Software Dependencies No The paper states in its checklist that training details and hyperparameters are included in the Appendix ('See the Appendix'), but the Appendix content is not provided for analysis in this excerpt. The main body of the paper does not specify any software dependencies with version numbers.
Experiment Setup No The paper states that 'More details on datasets, evaluation metrics, data pre-processing, experimental settings, network structures and their hyper-parameters are shown in the Appendix.' (Section 4). However, the Appendix content is not provided for analysis in this excerpt, and the main text does not include specific hyperparameter values or detailed training configurations.