This work rethink the ELBO under the Location-Scale Noise Model and finds that it naturally induces a Gaussian negative log likelihood objective for the estimators and inherently defines a joint training objective that unifies recent diffusion paradigms for probabilistic forecasting.
Abstract
Probabilistic time series forecasting requires modeling and predicting complex and time-varying distributions. Recently, Denoising Diffusion Probabilistic Model (DDPM)-based approaches have shown promise by equipping the dif- fusion process with pretrained mean and variance estimators to accommodate distributional shift. However, these methods typically follow the standard DDPM framework and consider only partial components of the evidence lower bound (ELBO), treating the training of estimators as designed regression tasks separate from the variational inference framework. To address this, we rethink the ELBO under the Location-Scale Noise Model (LSNM) and find that it naturally induces a Gaussian negative log likelihood objective for the estimators and inherently defines a joint training objective that unifies recent diffusion paradigms for probabilistic forecasting. Building on this principled ELBO reformulation, we propose Diff- PTS, a general framework that enables end-to-end optimization of all components within the ELBO. Across multiple benchmarks, DiffPTS consistently outperforms recent models, achieving state-of-the-art performance with an average CRPS/MSE reduction of over 14.53%/16.55% compared to existing diffusion-based methods. The code is available at https://github.com/wwy155/DiffPTS.
This study examines how forecast quality evolves throughout reverse diffusion to find that general temporal structure is often recovered at relatively high noise levels, whereas continued low-noise refinement can introduce statistical drift and degrade the final forecast.
Two-stage Odd Residual Flows (TORF), a framework that decouples mean forecasting from uncertainty estimation, achieves state-of-the-art deterministic accuracy (NMAE) while providing strong density estimation performance (CRPS) on short and long-horizon forecasting.
Kiran Madhusudhanan, Christian Klötergens, Lars Schmidt-Thieme et al.· 0 citations
Accurate probabilistic forecasting of multivariate time series (MTS) is crucial for many real-world applications but remains challenging due to the inherent complexity and non-stationarity of these signals. To address these challenges, we propose the Dynamic Decomposition Diffusion Model (D3M), which integrates advance...
Yu-Xin Li, Ao Wang, Xin-Yue Hu et al.· IEEE Transactions on Signal...· 0 citations
A stochastic few-step generator whose FID does not degrade as the sampling budget grows from 4 to 50 NFE, and the same recipe transfers to text-to-image generation.
Tommaso Martorella, Alexandre Galashov, F. Krause et al.· arXiv.org· 1 citation
This work proposes a methodology based on the standard ReLU Deep Neural Networks (DNN) to make predictions and quantify their uncertainty, and builds the so-called pertinent prediction interval (PPI) with the DNN model estimator.
DynG-Diff, a variable-sensitive dynamic guidance diffusion framework for probabilistic multivariate time-series forecasting, mathematically formulates this dynamic weight as the local precision of the observation distribution, enabling precise guidance for high-confidence variables during inference while filtering out...
Zhen-Te Zhang, Zheng-Wei Ni, Wei Fan· 0 citations
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduOct 13, 2026
Machine-learning networks can learn without detailed instructions from a designer. Those that do may offer insight into how the brain reorganizes itself during learning.
MIT News · Artificial Intelligence· news.mit.eduSep 29, 2026
Professor Sherry Turkle’s new book, “Artificial Intimacy,” offers a withering critique of chatbots and the antisocial dynamics she believes they encourage.