AI Paper Atlas

Research map / Time series / Foundation models for time series

Contrastive Learning for Unsupervised Domain Adaptation of Time Series

Yilmazcan Ozyurt, Stefan Feuerriegel, Ce Zhang

ICLR 2023 · Time series › Foundation models for time series · approach: Time series representation learning · 25 citations (OpenAlex)

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 changedWhereMeasuredEffect
Add contrastive learning (MoCo) and nearest-neighbor CL alignment on top of adversarial training (full CLUDA vs. w/o CL and w/o NNCL) (ablation)domain adversarial base architecture · WISDM, HAR, HHAR benchmarksaccuracyImproved
Remove the discriminator from CLUDA (w/o Discriminator) (ablation)CLUDA · benchmark datasetsaccuracy: discriminator gives consistent but smaller gainsImproved
Replace MoCo-based CL with SimCLR or NCL in CLUDA (ablation)CLUDA · benchmark datasetsaccuracyHurt
CLUDA vs. best baseline AdvSKM on MIMIC mortalityMIMIC-IV mortality UDA · MIMIC-IV 12 scenariosAUROC: +2.2% relative (0.773 vs 0.756)Improved
Use CLUDA instead of no-UDA on AUMC decompensationAUMC decompensation UDA · MIMIC to AUMCAUROC: fills 47.6% of performance gap vs 16.7% for best baseline HoMMImproved
Most UDA baselines (CAN, CDAN, DDC, DeepCORAL, DSAN, HoMM, MMDA, CoDATS) applied to MIMIC mortalityMIMIC-IV mortality UDA · MIMIC-IV 12 scenariosAUROC (avg): CDAN 0.721 below w/o UDA 0.750; VRADA 0.745 and AdvSKM 0.756 vary by scenarioMixed

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy