Conditional diffusion transformer enabled few-shot spatiotemporal modeling

Zhenzhen Zhao

doi:10.1016/j.inffus.2026.104499

Conditional diffusion transformer enabled few-shot spatiotemporal modeling

2026-05-27

0

PRE

AI

DOI：10.1016/j.inffus.2026.104499

Original

Original request for help

Save

Abstract

En 中文

• CDT: A few-shot spatiotemporal fusion modeling transformer framework. • Conditional diffusion process extracts dynamics in spatiotemporal data. • Construct the noise predictor with hierarchical spatiotemporal information. • Dynamically fuse diffusion-generated features and task-specific features. • The experiment results demonstrate the effectiveness of the framework.

Keywords:

Spatiotemporal features fusion

Conditional diffusion

Pre-training fine-tuning paradigm

Few-shot modeling

Conditional diffusion transformer enabled few-shot spatiotemporal modeling

Abstract

Journal

Organization