Reproduce “Deterministic Prediction as Global Velocity: Flow Matching for Probabilistic Time-Series Forecast” 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
Jiajing Hu, Xu Han, Li-Ping Liu
DV-Flow produces probabilistic time-series forecasts by combining a context-based point forecast as a global velocity with a learned, state- and time-dependent local velocity that transports noise into future trajectories. An auxiliary point-wise loss trains the deterministic predictor. In controlled tests, it achieved lower CRPS_sum than matched alternatives on all six datasets and showed more stable observed gradients, with added training and inference cost.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.