Meta-Learning Dynamics Forecasting Using Task Inference

TL;DR

This paper introduces DyAd, a meta-learning approach for dynamics forecasting that generalizes across different domains by inferring task-specific features and adapting the forecaster, outperforming existing methods on fluid and ocean data.

Contribution

The paper presents DyAd, a novel meta-learning model that partitions heterogeneous domains into tasks and adapts to new domains using task inference and domain generalization techniques.

Findings

DyAd outperforms state-of-the-art models on turbulent flow data.

DyAd effectively generalizes to real-world ocean data.

Theoretical analysis links generalization error to task relatedness and domain differences.

Abstract

Current deep learning models for dynamics forecasting struggle with generalization. They can only forecast in a specific domain and fail when applied to systems with different parameters, external forces, or boundary conditions. We propose a model-based meta-learning method called DyAd which can generalize across heterogeneous domains by partitioning them into different tasks. DyAd has two parts: an encoder which infers the time-invariant hidden features of the task with weak supervision, and a forecaster which learns the shared dynamics of the entire domain. The encoder adapts and controls the forecaster during inference using adaptive instance normalization and adaptive padding. Theoretically, we prove that the generalization error of such procedure is related to the task relatedness in the source domain, as well as the domain differences between source and target. Experimentally, we demonstrate that our model outperforms state-of-the-art approaches on both turbulent flow and real-world ocean data forecasting tasks.