Get Started
Home
Topics
Search
Library
Research questionHow can lightweight time-series forecasters learn accurate, specialized forecasts from scarce private data?Compact forecasters are cheaper to deploy than foundation models but usually need substantial task-specific training data. In many domains, observations arrive slowly or cannot be broadly shared, leaving too little data to specialize a model.
AI
AI Agents
Machine Learning
Multi-agent Systems
Small / On-device Models
Latest papersRecent research connected to this question, newest first.MetaCaster: Meta-Harness-Optimized Agent for End-to-End Few-Shot Learning of Lightweight Time Series ForecastersThe source studies lightweight time-series forecasters trained from a few examples and textual contexts, using agentic data generation. Evidence comes from experiments across 18 datasets, 23 lightweight forecasters, and 14 baselines; the reported results address data efficiency, computational efficiency, and forecasting quality.research paper · Sep 3, 2026
Related questions
How can online time-series forecasters adapt to evolving, unseen patterns without costly continual parameter updates?How can multimodal time-series forecasting align sparse text with heterogeneous temporal dynamics?How can recurrent forecasting models use fewer trainable parameters while preserving competitive performance and scalability?How can synthetic time-series generation preserve temporal and frequency structure when training data are scarce?