Research map / Time series / Foundation models for time series
Unsupervised Scalable Representation Learning for Multivariate Time Series
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 multiple negative samples rather than K=1 (ablation) | Unsupervised time-series encoder · UCR archive | classification accuracy | Helped |
| Change the number K of negative samples among values greater than 1 (ablation) | Unsupervised time-series encoder · UCR archive | classification accuracy | No clear effect |
| Combine representations from encoders trained with different K values (ablation) | Unsupervised time-series encoder · UCR classification | accuracy | Mixed |
| Use learned representations instead of raw values | Linear regressor · IHEPC Day | Test MSE: 8.92×10^{-2} -> 8.92×10^{-2} | No clear effect |
| Use learned representations instead of raw values | Linear regressor · IHEPC Day | Wall time: 3min 1s -> 12s | Helped |
| Use learned representations instead of raw values | Linear regressor · IHEPC Quarter | Test MSE: 6.26×10^{-2} -> 7.26×10^{-2} | Hurt |
| Use learned representations instead of raw values | Linear regressor · IHEPC Quarter | Wall time: 1h 40min 15s -> 9s | Helped |
| Use the learned representation space rather than raw-series DTW distances | One-nearest-neighbor classifier · UCR datasets | classification accuracy: 66% | Helped |
| Use the unsupervised representations with a small labeled subset rather than train ResNet on that subset | SVM and ResNet · TwoPatterns | accuracy: 26% -> 81% | Helped |
| Train the encoder on another dataset and transfer its representations | Unsupervised time-series encoder · UCR classification | accuracy | No clear effect |
| Replace the dilated convolutional encoder with an LSTM encoder (ablation) | Proposed unsupervised training procedure · Time-series representation learning | classification accuracy | Mixed |
Closest papers by meaning
- T-Rep: Representation Learning for Time Series using Time-Embeddings (ICLR 2024 · similarity 0.89)
- Time Series Representations with Hard-Coded Invariances (ICML 2025 · similarity 0.88)
- Learning Without Augmenting: Unsupervised Time Series Representation Learning via Frame Projections (NeurIPS 2025 · similarity 0.86)
- TSLANet: Rethinking Transformers for Time Series Representation Learning (ICML 2024 · similarity 0.86)
- TimeCHEAT: A Channel Harmony Strategy for Irregularly Sampled Multivariate Time Series Analysis (AAAI 2025 · similarity 0.86)
- Frequency-Masked Embedding Inference: A Non-Contrastive Approach for Time Series Representation Learning (AAAI 2025 · similarity 0.86)
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.