Get Started
Home
Topics
Search
Library
Research questionHow can time-series representations transfer across multiple source datasets while capturing dynamics relevant to the target task?Simple recovery objectives may not capture the temporal dynamics needed by downstream tasks. Transfer from only one source dataset can also leave useful cross-dataset information unused.
AI
Machine Learning
Research Paper
Statistical Machine Learning
Technology
Latest papersRecent research connected to this question, newest first.SMart: A Multi-source Multi-phase Time Series Representation Transfer FrameworkThe source addresses self-supervised time-series representation learning with source-to-target transfer. It reports results for classification and regression on univariate and multivariate datasets, including improvements in mean absolute error and accuracy, but does not establish applicability beyond the evaluated datasets and tasks.research paper · Sep 2, 2026
Related questions
How can multimodal time-series forecasting align sparse text with heterogeneous temporal dynamics?How can representation-learning methods for dynamic heterogeneous graphs be compared across temporal granularities?How can synthetic time-series generation preserve temporal and frequency structure when training data are scarce?How can synthetic medical time-series generation preserve rare-class patterns across temporal scales for downstream prediction?