![[2602.17706] Parallel Complex Diffusion for Scalable Time Series Generation](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.17706] Parallel Complex Diffusion for Scalable Time Series Generation
Summary
The paper presents PaCoDi, a novel approach to time series generation using parallel complex diffusion, enhancing efficiency and quality in modeling long-range dependencies.
Why It Matters
This research addresses critical limitations in traditional time series generation models, particularly regarding computational efficiency and representational capacity. By introducing a new architecture that operates in the frequency domain, it provides a scalable solution that could significantly impact various applications in machine learning and data analysis.
Key Takeaways
- PaCoDi decouples generative modeling in the frequency domain, enhancing computational efficiency.
- The model achieves a 50% reduction in attention FLOPs without loss of information.
- Extensive experiments demonstrate superior generation quality and inference speed compared to existing models.
- Theoretical foundations include Quadrature Forward Diffusion and Conditional Reverse Factorization theorems.
- The approach effectively handles non-isotropic noise distributions through a novel Heteroscedastic Loss.
Computer Science > Machine Learning arXiv:2602.17706 (cs) [Submitted on 10 Feb 2026] Title:Parallel Complex Diffusion for Scalable Time Series Generation Authors:Rongyao Cai, Yuxi Wan, Kexin Zhang, Ming Jin, Zhiqiang Ge, Qingsong Wen, Yong Liu View a PDF of the paper titled Parallel Complex Diffusion for Scalable Time Series Generation, by Rongyao Cai and 6 other authors View PDF HTML (experimental) Abstract:Modeling long-range dependencies in time series generation poses a fundamental trade-off between representational capacity and computational efficiency. Traditional temporal diffusion models suffer from local entanglement and the $\mathcal{O}(L^2)$ cost of attention mechanisms. We address these limitations by introducing PaCoDi (Parallel Complex Diffusion), a spectral-native architecture that decouples generative modeling in the frequency domain. PaCoDi fundamentally alters the problem topology: the Fourier Transform acts as a diagonalizing operator, converting locally coupled temporal signals into globally decorrelated spectral components. Theoretically, we prove the Quadrature Forward Diffusion and Conditional Reverse Factorization theorem, demonstrating that the complex diffusion process can be split into independent real and imaginary branches. We bridge the gap between this decoupled theory and data reality using a \textbf{Mean Field Theory (MFT) approximation} reinforced by an interactive correction mechanism. Furthermore, we generalize this discrete DDPM to cont...
