Seq-Flow improves forecast updates using fewer sampling steps
Seq-Flow: Efficient Probabilistic Forecasting with Self-Rollout Error Control
Machine Learning
Summary
Forecasting how something will change over time can be hard because predictions often need to be updated with new information. The authors propose Seq-Flow, a method that improves these updates by transforming old predictions into new ones directly, making it faster and more accurate when only a few update steps are used. They also developed a training approach that helps the method avoid building up errors over many updates. Experiments show Seq-Flow can give better predictions in scientific tasks like particle accelerator and fluid dynamics forecasting, even when making hundreds of updates.
What this means in practice
- •For particle accelerator operators: Improve near real-time forecasting of beam spill distributions to enhance control and safety during accelerator operation.
- •For fluid dynamics engineers: Generate faster and accurate updated predictions of fluid flow conditions to support system monitoring and design decisions.
Authors
Yinan Huang, Shitij Govil, Bo Dai, Pan Li
Abstract
Many scientific forecasting tasks require updating a distribution over future trajectories as new observations arrive. Conventional diffusion and flow models generate each forecast from Gaussian noise, often at the cost of many sampling steps. Warm-start methods reuse earlier predictions to reduce this cost, but their models are not trained to perform the forecast update itself, which can compromise quality under few-step sampling. In this work, we introduce Seq-Flow, a conditional flow model whose ODE transports samples from the previous forecast distribution to the updated one. Because successive forecasts often differ only modestly, this transport starts from an informative distribution and can produce accurate updates with few flow evaluations. Recursive reuse also creates a challenge: errors in one forecast become errors in the initial states of subsequent flows. We address this with self-rollout training, in which a moving average copy of the model generates forecasts that initialize later training updates. Unlike self-forcing methods, which reuse generated outputs as conditioning context, Seq-Flow reuses them as the source of the next flow. Experiments On particle-accelerator beam spill forecasting show Seq-Flow reduces CRPS by 65% under a few-NFE sampling budget, while remaining competitive with strong baselines on fluid-dynamics forecasting tasks. Although trained on self-rollouts of at most four updates, Seq-Flow remains accurate over more than 400 consecutive updates. Our code is available at https://github.com/Graph-COM/Seq-Flow.