SwiftTS: A Swift Selection Framework for Time Series Pre-trained Models via Multi-task Meta-Learning

2025-10-27

AI Summary

Model selection for time-series pre-trained models is time-consuming and requires individual fine-tuning per candidate model. Method: This paper proposes SwiftTS, the first framework to formulate model selection as a data–model compatibility prediction task. It employs a lightweight dual-encoder architecture to jointly encode time-series data features and pre-trained model representations. A patchwise compatibility scoring mechanism and a horizon-adaptive expert ensemble module are introduced to ensure cross-dataset and cross-horizon generalization, as well as out-of-distribution (OOD) robustness. Training is optimized via multi-task meta-learning and cross-dataset/cross-horizon sampling. Contribution/Results: Evaluated on 14 downstream datasets and 8 pre-trained models, SwiftTS achieves state-of-the-art performance—improving recommendation accuracy while reducing model selection time by over 90% compared to existing baselines.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Search and Optimization: Metareasoning and Metaheuristics
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Pre-trained models exhibit strong generalization to various downstream tasks. However, given the numerous models available in the model hub, identifying the most suitable one by individually fine-tuning is time-consuming. In this paper, we propose extbf{SwiftTS}, a swift selection framework for time series pre-trained models. To avoid expensive forward propagation through all candidates, SwiftTS adopts a learning-guided approach that leverages historical dataset-model performance pairs across diverse horizons to predict model performance on unseen datasets. It employs a lightweight dual-encoder architecture that embeds time series and candidate models with rich characteristics, computing patchwise compatibility scores between data and model embeddings for efficient selection. To further enhance the generalization across datasets and horizons, we introduce a horizon-adaptive expert composition module that dynamically adjusts expert weights, and the transferable cross-task learning with cross-dataset and cross-horizon task sampling to enhance out-of-distribution (OOD) robustness. Extensive experiments on 14 downstream datasets and 8 pre-trained models demonstrate that SwiftTS achieves state-of-the-art performance in time series pre-trained model selection.

Problem

• SwiftTS rapidly selects optimal pre-trained models for time series tasks
• It predicts model performance without costly fine-tuning of all candidates
• The framework enhances generalization across datasets and horizons

Innovation

• Learning-guided approach predicts model performance
• Lightweight dual-encoder embeds time series and models
• Horizon-adaptive expert composition enhances generalization robustness