The paper introduces SCROLL, a method for forecasting multiple observables in stochastic dynamical systems by composing each observable’s likelihood into per‑task free‑routed last‑layer beliefs on a shared backbone. This approach learns unit‑dependent loss scaling directly from data, enabling accurate predictive variance estimation without separate tuning. Experiments on the Ornstein–Uhlenbeck process, stochastic Lorenz‑63, and real air‑quality data show that SCROLL recovers analytic kernels, achieves superior negative log‑likelihood on state and regime tasks, and maintains calibration while reducing hyper‑parameter search costs.
arXiv:2609.08412v1 Announce Type: new Abstract: Machine-learning weather models (MLWMs) now match or outperform operational numerical weather prediction (NWP) at global medium-range forecasting, at f...
arXiv:2608.30795v1 Announce Type: cross Abstract: End-to-end weather forecasting systems produce skillful global gridded and station forecasts directly from raw Earth observations, replacing the nume...
arXiv:2606. 04342v1 Announce Type: cross Abstract: Multi-step time series forecasting (MSF) is commonly evaluated using point-wise error metrics such as mean squared error (MSE), implicitly treating the conditional mean as a sufficient target.
The paper introduces DIAL, a method that dynamically adjusts the native stochasticity of a frozen weather forecasting model to improve ensemble reliability across variables, scales, and lead times. DIAL learns from past forecasts to predict necessary spread adjustments and applies small noise updates during each forecast step, avoiding retraining or full backpropagation. Experiments on FourCastNet 3 show that DIAL outperforms post‑processing and reward‑based baselines, achieving an 8.27% improvement in multiscale CRPS over the unmodified model.
The paper introduces a new diffusion-based approach for probabilistic weather forecasting that incorporates auxiliary conditional denoising tasks. By predicting future states from both clean and corrupted contexts, the model reduces ambiguity and learns conditional distributions across noise levels using a proper scoring rule. Experiments demonstrate improved one‑step forecasting, especially at longer horizons, and better calibration in high‑dimensional global weather scenarios.
arXiv:2606. 27711v1 Announce Type: cross Abstract: We introduce a neural network-based framework for learning time series estimators through a process we term decision-theoretic pretraining.
arXiv:2607. 10362v1 Announce Type: new Abstract: Latent world models are trained to predict future states in a learned representation and are then deployed inside a planner that selects actions by simulating them forward.
arXiv:2609.09586v1 Announce Type: new Abstract: Time series foundation models (TSFMs) are increasingly pre-trained on synthetically generated time series trajectories, where the data generating proce...
arXiv:2610.09770v1 Announce Type: cross Abstract: Deep learning has made rapid advances in weather forecasting: autoregressive models trained on atmospheric reanalyses now rival dynamical models acro...
The paper introduces Internal Dual-Wiener routing (Internal‑DW), a backward‑only method that weight‑balances internal gradient routes in autoregressive forecasting. By estimating bounded Wiener gains for identity and nonlinear paths, it suppresses unpredictable noise while preserving predictable learning signals, reducing forecast error by 5.2%–13.8% on four weak‑drive testbeds compared to full BPTT and outperforming gradient clipping, Jacobian regularization, and truncated BPTT in most cases. The approach shows that long‑horizon supervision can be effective without trusting every backward gradient equally.
arXiv:2605. 05540v2 Announce Type: replace Abstract: Fast surrogate modeling for high-dimensional physical dynamics requires more than low short-term error: useful models must roll out efficiently while preserving the statistical structure of long trajectories.
Hybrid forecasting systems that combine machine learning with numerical weather prediction are explored using a neural‑network representation of model error within the Integrated Forecasting System. The study introduces a targeted neural‑network 4D‑Var approach that estimates only network parameters and a batched extension that processes assimilation windows in random order, averaging parameter updates across multiple windows. Experiments show that while sequential online learning improves forecasts during training, it fails to generalise, whereas the batched formulation consistently enhances out‑of‑sample forecast skill and improves convergence with larger batch sizes.