DYNA-PRUNER: Input-Adaptive Data-Model Co-Pruning for Efficient and Scalable Spatio-Temporal Media Prediction

2026-06-13

AI Summary

This work addresses the high computational cost and limited real-time deployability of existing spatiotemporal forecasting models, which often overlook dynamic redundancy in inputs—such as clear skies or calm seas. To overcome these limitations, we propose Dyna-Pruner, a novel framework featuring a shared importance synchronization mechanism that enables input-adaptive cooperative pruning. Dyna-Pruner jointly generates coupled masks for both data regions and computational units, dynamically constructing sparse subnetworks tailored to each sample during inference. The approach is compatible with diverse backbone architectures—including CNNs, RNNs, and Transformers—and supports end-to-end efficient spatiotemporal prediction. Experiments on WeatherBench, SEVIR, and TaxiBJ benchmarks demonstrate up to 70% FLOPs reduction, 2.5× speedup on Jetson AGX Orin hardware, and less than 1% accuracy degradation.

Abstract

Spatio-temporal prediction supports radar/satellite nowcasting and city-scale traffic monitoring, but modern models are often too expensive for real-time deployment. This stems from a mismatch between dense computation and strong input-dependent redundancy (e.g., calm seas or clear skies). To enable automated, resource-aware architecture optimization in scalable media analysis, we propose Dyna-Pruner, an end-to-end framework for input-dependent co-pruning of data and model structure. A shared-importance synchronization mechanism generates coupled masks that prune redundant regions and their corresponding computational units (e.g., convolutional filters), yielding per-sample sparse sub-networks at inference time. Experiments on WeatherBench, SEVIR, and TaxiBJ show seamless integration with CNN, RNN, and Transformer backbones, reducing FLOPs by up to $70\%$ and achieving a $2.5\times$ speedup on NVIDIA Jetson AGX Orin with negligible accuracy loss ($<1\%$).

Problem

• spatio-temporal prediction
• model redundancy
• real-time deployment
• input-dependent redundancy
• computational efficiency

Innovation

• input-adaptive pruning
• data-model co-pruning
• spatio-temporal prediction
• dynamic sparse sub-networks
• shared-importance synchronization