Research map / Time series / Foundation models for time series
Time Series Contrastive Learning with Information-Aware Augmentations
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| use information-aware adaptive augmentations instead of TS2Vec's universal augmentation | Univariate forecasting representations · Univariate forecasting, average across datasets and horizons | MSE: 0.175 -> 0.154 | Helped |
| use information-aware adaptive augmentations instead of TS2Vec's universal augmentation | Multivariate forecasting representations · Multivariate forecasting, average across datasets and horizons | MSE: 0.852 -> 0.805 | Helped |
| use supervised labels to guide the augmentation meta-learner | InfoTS · UEA 30 datasets | Avg. ACC: 0.714 -> 0.730 | Helped |
| randomly select a candidate augmentation instead of adaptively selecting it (ablation) | InfoTS · Electricity univariate forecasting | MSE: 0.370 -> 0.384 | Hurt |
| apply all candidate transformations sequentially instead of adaptively selecting them (ablation) | InfoTS · Electricity univariate forecasting | MSE: 0.370 -> 0.382 | Hurt |
| remove the fidelity objective (ablation) | InfoTS meta-learner · Electricity univariate forecasting | MSE: 0.370 -> 0.384 | Hurt |
| remove the variety objective (ablation) | InfoTS meta-learner · Electricity univariate forecasting | MSE: 0.370 -> 0.375 | Hurt |
| add information-aware adaptive augmentation to the TS2Vec backbone (ablation) | TS2Vec · Electricity univariate forecasting | MSE: 0.393 -> 0.380 | Helped |
| use subsequence augmentation only instead of adaptive augmentation selection (ablation) | InfoTS · Electricity univariate forecasting | MSE: 0.370 -> 0.374 | Hurt |
| use subsequence augmentation only instead of adaptive augmentation selection (ablation) | InfoTS · ETTh1 univariate forecasting | MSE: 0.091 -> 0.105 | Hurt |
| use a single fixed candidate transformation rather than adaptive combinations (ablation) | InfoTS · ETTh1 univariate forecasting | MSE: 0.091 vs 0.098; 0.099; 0.106; 0.101; 0.103; 0.105 | Mixed |
| use a single fixed candidate transformation rather than adaptive combinations (ablation) | InfoTS · Electricity univariate forecasting | MSE: 0.369 vs 0.374; 0.377; 0.381; 0.383 | Hurt |
Closest papers by meaning
- Parametric Augmentation for Time Series Contrastive Learning (ICLR 2024 · similarity 0.94)
- Inference via Interpolation: Contrastive Representations Provably Enable Planning and Inference (NeurIPS 2024 · similarity 0.88)
- Soft Contrastive Learning for Time Series (ICLR 2024 · similarity 0.88)
- Self-Supervised Contrastive Learning for Long-term Forecasting (ICLR 2024 · similarity 0.87)
- Contrastive Difference Predictive Coding (ICLR 2024 · similarity 0.87)
- A Statistical Theory of Contrastive Learning via Approximate Sufficient Statistics (NeurIPS 2025 · similarity 0.87)
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.