Research map / Time series / Foundation models for time series
AutoDA-Timeseries: Automated Data Augmentation for Time Series
Zijun Dou, Zhenhe Yao, Zhe Xie, Xidao Wen, Tong Xiao, Dan Pei
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add time series features to condition augmentation policy (ablation) | AutoDA-Timeseries (RNN) · Long-term forecasting, Exchange | MSE: 0.4301 -> 0.3409 | Improved |
| add time series features (ablation) | AutoDA-Timeseries (RNN) · Long-term forecasting, ETTh1 | MSE: 0.5612 -> 0.5429 | Improved |
| add time series features (ablation) | AutoDA-Timeseries (RNN) · Long-term forecasting, Weather | MSE: 0.2689 -> 0.2574 | Improved |
| add joint optimization of probabilities and intensities (ablation) | AutoDA-Timeseries (RNN) · Long-term forecasting, Exchange | MSE: 0.4395 -> 0.3409 | Improved |
| add dynamic temperature (ablation) | AutoDA-Timeseries (RNN) · Long-term forecasting, Exchange | MSE: 0.4235 -> 0.3409 | Improved |
| add composite loss (ablation) | AutoDA-Timeseries (RNN) · Long-term forecasting, Exchange | MSE: 0.4322 -> 0.3409 | Improved |
| include autocorrelation Catch22 features (ablation) | AutoDA-Timeseries · UEA classification EP | accuracy: 0.9275 -> 0.9783 | Improved |
| include differencing Catch22 features (ablation) | AutoDA-Timeseries · UEA classification RS | accuracy: 0.8026 -> 0.8947 | Improved |
| initialize with uniform distribution plus raw bonus instead of pure uniform (ablation) | AutoDA-Timeseries · UEA classification (ROCKET) | average accuracy: 0.9187 -> 0.9353 | Improved |
| apply UniformAugment instead of no augmentation in cross-dataset transfer | NoAug · ETTh1 to ETTh2 transfer (RNN) | MSE: 0.5290 -> 0.7188 | Hurt |
| apply AutoDA-Timeseries instead of no augmentation in cross-dataset transfer | NoAug · ETTh1 to ETTh2 transfer (RNN) | MSE: 0.5290 -> 0.5076 | Improved |
| apply TS2Vec representation learning instead of no augmentation | NoAug · UEA classification (TCN) | accuracy: 0.685 -> 0.584 | Hurt |
Closest papers by meaning
- Parametric Augmentation for Time Series Contrastive Learning (ICLR 2024 · similarity 0.89)
- Large Pre-trained time series models for cross-domain Time series analysis tasks (NeurIPS 2024 · similarity 0.85)
- TEMPO: Prompt-based Generative Pre-trained Transformer for Time Series Forecasting (ICLR 2024 · similarity 0.85)
- TimeDART: A Diffusion Autoregressive Transformer for Self-Supervised Time Series Representation (ICML 2025 · similarity 0.85)
- Time Series Generation Under Data Scarcity: A Unified Generative Modeling Approach (NeurIPS 2025 · similarity 0.85)
- T-Rep: Representation Learning for Time Series using Time-Embeddings (ICLR 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
