Accelerating Frequency Domain Diffusion Models with Error-Feedback Event-Driven Caching

2026-04-24

AI Summary

Diffusion models suffer from low inference efficiency in time series generation, hindering their practical deployment. To address this, this work proposes the E²-CRF method, which leverages frequency-domain modeling—exploiting spectral locality and mirror symmetry—to reduce computational dimensionality. It further introduces an event-driven residual dynamic triggering mechanism and a closed-loop error feedback caching strategy to adaptively reuse Transformer key-value (KV) features, aligning with the diffusion process’s progression from coarse structure to fine details. Evaluated on five datasets, the approach achieves approximately 2.2× faster inference while preserving generation quality, and it provides theoretical guarantees on both approximation error bounds and computational complexity.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines

Abstract

Diffusion models achieve remarkable success in time series generation. However, slow inference limits their practical deployment. We propose E$^2$-CRF (Error-Feedback Event-Driven Cumulative Residual Feature caching) to accelerate frequency domain diffusion models. Our method exploits two structural properties: (1) spectral localization, where signal energy concentrates in low frequencies, and (2) mirror symmetry, which halves the effective frequency dimension. E$^2$-CRF uses a closed-loop error-feedback system that adaptively caches transformer KV features across diffusion steps. We trigger recomputation using event-driven residual dynamics instead of fixed schedules. Our method selectively recomputes high-energy or rapidly-changing tokens while reusing cached features for stable high-frequency components. E$^2$-CRF achieves ~2.2 speedup while maintaining sample quality. We demonstrate effectiveness on 5 datasets. Our caching strategy naturally aligns with the diffusion process's structure-to-detail progression. We include sufficient-condition error and complexity bounds under standard regularity assumptions (Appendix), alongside empirical validation. Our code is available at https://github.com/NoakLiu/FastFourierDiffusion and is also integrated in https://github.com/NoakLiu/FastCache-xDiT.

Problem

• diffusion models
• time series generation
• inference acceleration
• frequency domain

Innovation

• frequency domain diffusion
• error-feedback caching
• event-driven recomputation
• spectral localization
• KV feature reuse