Meta-forecasting is a newly emerging field which combines meta-learning and time series forecasting. The goal of meta-forecasting is to train over a collection of source time series and generalize to new time series one-at-a-time. Previous approaches in meta-forecasting achieve competitive performance, but with the restriction of training a separate model for each sampling frequency. In this work, we investigate meta-forecasting over different sampling frequencies, and introduce a new model, the Continuous Frequency Adapter (CFA), specifically designed to learn frequency-invariant representations. We find that CFA greatly improves performance when generalizing to unseen frequencies, providing a first step towards forecasting over larger multi-frequency datasets.
@article{arxiv.2302.02077,
title = {Cross-Frequency Time Series Meta-Forecasting},
author = {Mike Van Ness and Huibin Shen and Hao Wang and Xiaoyong Jin and Danielle C. Maddix and Karthick Gopalswamy},
journal= {arXiv preprint arXiv:2302.02077},
year = {2023}
}