Reproduce “Quantifying the Sampling Frequency Invariance of Time-Series Foundation Models for Sensor Data Applications” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Sourish Gunesh Dhekane, Abinaya Kumar, Torsten Woertwein, Yan-Bo Lin, Thomas Ploetz, Shabnam Ghaffarzadegan
The authors test four transformer-based time-series foundation models under reduced sampling frequency by band-limiting fixed-length sensor windows while preserving their length. They compare embedding drift, forecasting, and classification degradation. Classification degradation depends more on dataset configuration than model on average: MOMENT and TimesFM have mean CDI values of 0.112 and 0.114. All four models collapse in a near-critical bearing setting.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.