A Decomposable Forward Process in Diffusion Models for Time-Series Forecasting
This work addresses the challenge of maintaining long-term patterns in time-series forecasting for applications in fields like finance or climate, representing an incremental improvement by modifying only the diffusion process while remaining compatible with existing models.
The paper tackles the problem of preserving structured temporal patterns like seasonality in time-series forecasting by introducing a model-agnostic forward diffusion process that decomposes signals into spectral components, resulting in improved forecast quality across standard benchmarks with negligible computational overhead.
We introduce a model-agnostic forward diffusion process for time-series forecasting that decomposes signals into spectral components, preserving structured temporal patterns such as seasonality more effectively than standard diffusion. Unlike prior work that modifies the network architecture or diffuses directly in the frequency domain, our proposed method alters only the diffusion process itself, making it compatible with existing diffusion backbones (e.g., DiffWave, TimeGrad, CSDI). By staging noise injection according to component energy, it maintains high signal-to-noise ratios for dominant frequencies throughout the diffusion trajectory, thereby improving the recoverability of long-term patterns. This strategy enables the model to maintain the signal structure for a longer period in the forward process, leading to improved forecast quality. Across standard forecasting benchmarks, we show that applying spectral decomposition strategies, such as the Fourier or Wavelet transform, consistently improves upon diffusion models using the baseline forward process, with negligible computational overhead. The code for this paper is available at https://anonymous.4open.science/r/D-FDP-4A29.