Research map / Time series / Foundation models for time series
Time-Frequency Augmented Multi-level Contrastive Clustering for Time Series
Congyu Wang, Mingjing Du, Xiang Jiang
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 temporal-level contrastive loss (ablation) | TFMCC · UCR (40 datasets) | average NMI: 0.591 -> 0.632 | Improved |
| add instance-level contrastive loss (ablation) | TFMCC · UCR (40 datasets) | average NMI: 0.545 -> 0.632 | Improved |
| add cluster-level contrastive loss instead of k-means (ablation) | TFMCC · UCR (40 datasets) | average NMI: 0.584 -> 0.632 | Improved |
| add multi-scale time-domain augmentation (ablation) | TFMCC · UCR (40 datasets) | average NMI: 0.565 -> 0.632 | Improved |
| add multi-scale frequency-domain augmentation (ablation) | TFMCC · UCR (40 datasets) | average NMI: 0.566 -> 0.632 | Improved |
| add multi-scale time-frequency augmentation (both domains) (ablation) | TFMCC · UCR (40 datasets) | average NMI: 0.545 -> 0.632 | Improved |
| use multi-scale cropping instead of fixed cropping ratio in time domain (ablation) | TFMCC · UCR (40 datasets) | average NMI | Improved |
| use multi-scale DWT levels instead of fixed level in frequency domain (ablation) | TFMCC · UCR (40 datasets) | average NMI | Improved |
| vary temperature coefficients (ablation) | TFMCC · FaceAll, FacesUCR, SmoothSubspace, SyntheticControl | clustering performance | No clear effect |
| joint end-to-end training with clustering instead of two-stage k-means (TFMCC vs TS2Vec) | TS2Vec · UCR (40 datasets) | average NMI: 0.464 -> 0.632 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- TimesURL — TimesURL: Self-Supervised Contrastive Learning for Universal Time Series Representation Learning (AAAI 2024) · better on 4 of 4 shared results
- TS2Vec — TS2Vec: Towards Universal Representation of Time Series (AAAI 2022) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Cross-Domain Contrastive Learning for Time Series Clustering (AAAI 2024 · similarity 0.91)
- Parametric Augmentation for Time Series Contrastive Learning (ICLR 2024 · similarity 0.85)
- Frequency-Masked Embedding Inference: A Non-Contrastive Approach for Time Series Representation Learning (AAAI 2025 · similarity 0.85)
- MF-CLR: Multi-Frequency Contrastive Learning Representation for Time Series (ICML 2024 · similarity 0.84)
- Super Deep Contrastive Information Bottleneck for Multi-modal Clustering (ICML 2025 · similarity 0.84)
- TimesURL: Self-Supervised Contrastive Learning for Universal Time Series Representation Learning (AAAI 2024 · similarity 0.84)
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.
