New Millisecond-Resolution Dataset Expands Time Series Foundation Models to High-Frequency Wireless Networks
A groundbreaking dataset has been launched to enhance time series foundation models (TSFMs) by providing millisecond-resolution data on wireless and traffic conditions from an active 5G deployment. Existing large-scale datasets mainly concentrate on low-frequency time series, with sampling intervals of seconds to years, which limits their effectiveness in capturing high-frequency data intricacies. This new resource broadens the application of TSFMs by integrating high-frequency data for pre-training, adding wireless networks to the mix alongside established fields like energy and finance. It offers use cases for short-term forecasting with prediction horizons ranging from 1 millisecond to 96 milliseconds. The dataset, identified as arXiv:2603.16497v2, was released under the announcement type of replace-cross.
Key facts
- Dataset captures millisecond-resolution wireless and traffic conditions from operational 5G deployment
- Addresses limitation where current datasets focus on low-frequency time series (seconds to years)
- Expands time series foundation models to incorporate high-frequency data for pre-training
- Introduces wireless networks as new domain complementing energy and finance
- Provides use cases for short-term forecasting with 1-96 millisecond prediction horizons
- Announced as arXiv:2603.16497v2 with replace-cross announcement type
- Enables TSFMs to adapt across varying domains and temporal frequencies
- Bridges gap in high-frequency time series data availability
Entities
—