Frequency-Aligned Knowledge Distillation for Lightweight Spatiotemporal Forecasting
This work addresses the need for more efficient spatiotemporal forecasting models in domains like traffic and weather, though it is incremental as it builds on existing knowledge distillation techniques.
The paper tackles the problem of inefficient and memory-intensive spatiotemporal forecasting models by proposing a lightweight framework called Spectral Decoupled Knowledge Distillation (SDKD), which transfers multi-scale representations from a complex teacher to a student network, achieving reductions of up to 81.3% in MSE and 52.3% in MAE on the Navier-Stokes equation dataset.
Spatiotemporal forecasting tasks, such as traffic flow, combustion dynamics, and weather forecasting, often require complex models that suffer from low training efficiency and high memory consumption. This paper proposes a lightweight framework, Spectral Decoupled Knowledge Distillation (termed SDKD), which transfers the multi-scale spatiotemporal representations from a complex teacher model to a more efficient lightweight student network. The teacher model follows an encoder-latent evolution-decoder architecture, where its latent evolution module decouples high-frequency details and low-frequency trends using convolution and Transformer (global low-frequency modeler). However, the multi-layer convolution and deconvolution structures result in slow training and high memory usage. To address these issues, we propose a frequency-aligned knowledge distillation strategy, which extracts multi-scale spectral features from the teacher's latent space, including both high and low frequency components, to guide the lightweight student model in capturing both local fine-grained variations and global evolution patterns. Experimental results show that SDKD significantly improves performance, achieving reductions of up to 81.3% in MSE and in MAE 52.3% on the Navier-Stokes equation dataset. The framework effectively captures both high-frequency variations and long-term trends while reducing computational complexity. Our codes are available at https://github.com/itsnotacie/SDKD