New Shapelet Matching method improves time series transfer learning

Researchers have developed a new training-free method called Shapelet Matching for selecting source datasets in time series transfer learning. This approach identifies discriminative shapelets from target and potential source datasets to quantify similarity, aiming to mitigate negative transfer by combining multiple suitable sources for pre-training. Evaluations on the UCR Archive using CNN and Transformer architectures showed that multi-source pre-training reduces the risk of negative transfer and that Shapelet Matching performs strongly, particularly with CNN backbones, while avoiding the computational cost of full pre-training for each source. AI

IMPACT Improves efficiency and accuracy in time series classification by optimizing source dataset selection for transfer learning.

RANK_REASON The cluster contains a research paper detailing a new method for time series transfer learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX Code Finder for Papers
• Connected Papers
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• IArxiv Recommender
• Litmaps
• ScienceCast
• scite Smart Citations
• Shapelet Matching
• Transformer++
• UCR Archive

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →