Learning Deep Time-index Models for Time Series Forecasting

Authors: Gerald Woo, Chenghao Liu, Doyen Sahoo, Akshat Kumar, Steven Hoi

ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on real world datasets in the long sequence time-series forecasting setting demonstrate that our approach achieves competitive results with state-of-the-art methods, and is highly efficient. Code is available at https: //github.com/salesforce/Deep Time. We conduct extensive experiments on the long sequence time series forecasting (LSTF) setting, demonstrating Deep Time to be extremely competitive with state-of-the-art baselines. At the same time, Deep Time is highly efficient in terms of runtime and memory.
Researcher Affiliation Collaboration 1Salesforce Research Asia 2School of Computing and Information Systems, Singapore Management University. Correspondence to: Gerald Woo <gwoo@salesforce.com>, Chenghao Liu <chenghao.liu@salesforce.com>.
Pseudocode Yes Pseudocode of Deep Time is available in Appendix A. Algorithm 1 Py Torch-Style Pseudocode of Closed-Form Ridge Regressor. Algorithm 2 Py Torch-Style Pseudocode of Deep TIMe.
Open Source Code Yes Code is available at https: //github.com/salesforce/Deep Time.
Open Datasets Yes Experiments are performed on 6 real-world datasets Electricity Transformer Temperature (ETT), Electricity Consuming Load (ECL), Exchange, Traffic, Weather, and Influenzalike Illness (ILI) with full details in Appendix F. Appendix F states for ETT: 1https://github.com/zhouhaoyi/ETDataset.
Dataset Splits Yes The datasets are split into train, validation, and test sets chronologically, following a 70/10/20 split for all datasets except for ETTm2 which follows a 60/20/20 split, as per convention.
Hardware Specification Yes All experiments are performed on an Nvidia A100 GPU.
Software Dependencies No The paper mentions
Experiment Setup Yes Further implementation details on Deep Time are reported in Appendix G, and detailed hyperparameters are reported in Appendix H. Table 6 in Appendix H lists: Epochs 50, Learning rate 1e-3, Warm up epochs 5, Batch size 256, Early stopping patience 7, Max gradient norm 10.0, Layers 5, Layer size 256, λ initialization 0.0, Scales [0.01, 0.1, 1, 5, 10, 20, 50, 100], Fourier features size 4096, Dropout 0.1.