Physiome-ODE: A Benchmark for Irregularly Sampled Multivariate Time Series Forecasting Based on Biological ODEs
arXiv:2502.07489v2 Announce Type: replace Abstract: State-of-the-art methods for forecasting irregularly sampled time series with missing values predominantly rely on just four datasets and a few small toy examples for evaluation. While ordinary differential equations (ODE) are the prevalent models in science and engineering, a baseline model that forecasts a constant value outperforms ODE-based models from the last five years on three of these existing datasets. This unintuitive finding hampers further research on ODE-based models, a more plausible model family. In this paper, we develop a me
The proliferation of AI applications necessitates more robust and reliable methods for handling complex, real-world data, pushing the boundaries of current time-series forecasting. This new benchmark addresses known limitations in evaluating ODE-based models, which are critical for scientific and engineering domains.
Improved benchmarks for irregularly sampled multivariate time series forecasting are crucial for advancing AI's ability to model and predict complex systems, particularly those with underlying physical dynamics. This directly impacts the reliability and applicability of AI in critical scientific and engineering fields.
The introduction of Physiome-ODE provides a more challenging and biologically relevant benchmark, potentially shifting research efforts away from toy examples and towards more practical and performant ODE-based AI models. It highlights deficiencies in current forecasting methods and offers a path to overcome them.
- · AI researchers in time series forecasting
- · Biological and medical AI applications
- · Engineering and scientific modeling
- · Developers of robust AI models
- · Overly simplified time series forecasting models
- · AI applications reliant on current, less robust benchmarks
The benchmark reveals that current ODE-based models often underperform simpler baselines on existing datasets, indicating a gap in current AI forecasting capabilities.
This may lead to a renewed focus on developing more sophisticated and context-aware ODE-based AI models, particularly in fields where physical laws are paramount.
Successful development of these models could unlock breakthroughs in areas like drug discovery, personalized medicine, and climate modeling by enabling more accurate predictions from sparse or irregular data.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG