Quantifying the Sampling Frequency Invariance of Time-Series Foundation Models for Sensor Data Applications
Organizations: School of Interactive Computing, Georgia Institute of Technology, USA · Bosch Research, USA - Bosch Center for Artificial Intelligence
Abstract
Time-Series Foundation Models (TSFMs) are increasingly used in sensor applications, where sampling frequency (SF) varies across sensors, devices, and operating modes (e.g., power saving). Ideally, TSFMs must remain effective during such SF changes, provided the SF retains task-critical bandwidth. In this paper, we stress-test 4 prominent TSFMs: MOMENT, Chronos-2, Moirai-MoE, and TimesFM, to quantify their SF-invariance and investigate whether their behavior tracks the resulting loss of information bandwidth rather than the SF value itself. Our stress-test probes 3 dimensions: (i) the shift in embeddings, (ii) the degradation in forecasting performance, and (iii) the degradation in downstream performance, quantified by three novel metrics: FSI, FDI, and CDI, respectively. To our knowledge, this is the first study to place SF-invariance at the core of TSFM evaluations. Our results characterize: (i) how SF-invariance depends on the margin between signal's native-SF and the bandwidth its task requires, (ii) how much of that invariance the choice of TSFM explains, and (iii) how much of TSFM's performance degradation under a lower SF reflects information loss rather than a limitation of their abilities. These findings reframe SF-invariance as a measurable, signal-aware property of the (signal X TSFM) pair, and guide practitioners on when TSFMs display SF-invariance.
Figures & tables
| TSFM | Backbone | Majority Pre-train SF | Pre-train loss |
|---|---|---|---|
| MOMENT [ 5 ] | enc.-only (T5) | Hz | masked recon. (MSE) |
| Chronos-2 [ 1 ] | enc.-only (T5) | Hz | quantile (pinball) |
| Moirai-MoE [ 11 ] | dec.-only, MoE | Hz | distrib. NLL |
| TimesFM [ 4 ] | dec.-only | Hz | next-patch (MSE) |
| Datasets | HHAR | Bearing | MIT-BIH | |||||
|---|---|---|---|---|---|---|---|---|
| Dim | TSFM | Acc 100Hz | Acc 200Hz | Gyr 100Hz | Gyr 200Hz | Vib 12kHz | Vib 48kHz | ECG 125Hz |
| (1) FSI mean std | Chronos-2 | 0.061 0.059 | 0.068 0.045 | 0.026 0.016 | 0.033 0.021 | 0.092 0.023 | 0.027 0.017 | 0.125 0.064 |
| Moirai-MoE | 0.023 0.025 | 0.036 0.031 | 0.011 0.012 | 0.015 0.017 | 0.098 0.015 | 0.006 0.002 | 0.022 0.011 | |
| MOMENT | 0.023 0.026 | 0.028 0.033 | 0.008 0.009 | 0.004 0.007 | 0.116 0.026 | 0.006 0.001 | 0.012 0.006 | |
| TimesFM | 0.032 0.066 | 0.031 0.036 | 0.013 0.014 | 0.016 0.015 | 0.051 0.042 | 0.006 0.006 | 0.027 0.027 | |
| (2) FDI median (IQR) | Chronos-2 | 0.439 0.597 | 0.655 0.750 | 0.253 0.680 | 0.260 1.279 | 0.079 0.332 | 0.142 0.451 | 4.387 10.474 |
Time-Series Foundation Models
Time-series foundation models are pretrained sequence models designed to transfer across forecasting, classification, anomaly detection, and other temporal tasks.
See all 345 papers