Searcharxiv⌕ Search
Home/Archive/arXiv

arXiv · 2610.18390

Beyond RMSE and Granger Gain: Multi-Horizon Forecast-Error Entropy for Nonlinear Predictive Coupling

Andrei Velichko
View original source ↗Read source document ↗Export metadata ↓

Abstract

We investigate whether the past of an auxiliary time series provides predictive information about a target beyond that contained in the target's own history. We compare matched out-of-sample target-only and target-plus-source forecasts and introduce the forecast-error distribution entropy (FEDE) predictive-causality framework, where predictive causality is understood in the Granger sense: auxiliary history is informative when it improves prediction beyond the target's own past. FEDE quantifies how the full signed forecast-error distribution changes across prediction horizons. Controlled nonlinear benchmarks hide source-target dependence behind nonlinear transformations not supplied to the predictors. LightGBM recovers these hidden channels, whereas Ridge does not. Across four equal-power nonlinearities and four noise levels, the mean response-sensitivity statistic for FEDE remains 29.00, 27.47, 23.22, and 28.67, compared with 14.09, 10.68, 12.67, and 12.71 for mean log-RMSE gain. Directional and source controls separate the true source-to-target channel from the reverse direction and from independent periodic, quasiperiodic, chaotic, and autoregressive auxiliaries. Chronological Granger-style comparisons show strong model-form dependence: a degree-2 polynomial model is exceptionally sensitive to the matched quadratic channel but becomes weak or oppositely aligned for other nonlinearities, while basic convergent cross mapping is weak or directionally inconsistent in this stochastic, partly non-invertible setting. FEDE remains positively aligned across all 108 tested combinations of window length, history length, and maximum forecast horizon. Together, these results establish FEDE as a sensitive and robust horizon-resolved tool that extracts predictive information beyond conventional error-magnitude measures when the nonlinear interaction form is unknown. Code available in Zenodo.

Explore related subjects

nlin.CDcs.LG

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Andrei Velichko. 2026-10-13. Beyond RMSE and Granger Gain: Multi-Horizon Forecast-Error Entropy for Nonlinear Predictive Coupling. https://arxiv.org/abs/2610.18390

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

Intrinsic instantaneous coarse-to-fine recoverability in multiscale chaos: The Lorenz-96 system

In multiscale chaotic systems, a central question for closure is how much of an unresolved fine-scale target can be inferred from a simultaneous coarse observation. For a specified observation-target pair, we quantify instantaneous recoverability under the invariant snapshot measure as the fraction of target variance explained by conditioning on the retained coarse variables. Evaluating this quantity across scales produces a scale-resolved recoverability map. Applied to the Lorenz-96 system across a range of forcings, the maps reveal a strongly nonuniform landscape: lower-wavenumber targets have little variance explained by the retained lower modes, while a finite band of higher modes exhibits partial recoverability once the retained scales reach the energetically active intermediate range. Recoverability onset is organized around a reference cutoff at which two retained modes can directly couple to the target through the quadratic nonlinearity, although its precise location depends on attractor statistics. Across the tested forcings, this qualitative onset structure persists while recoverability decreases, leaving a larger fraction of target variance unexplained in more strongly driven regimes. This definition identifies the maximal explained-variance fraction attainable by instantaneous deterministic models under squared loss; the estimated maps provide a scale-dependent diagnostic for closure modeling.

nlin.CD↗
arXiv2026

Stability of inverse cascade in NLS Wave turbulence

We study the stability of the inverse cascade in the 3D Nonlinear Schrodinger Wave Kinetic Equation (WKE) in the presence and absence of a condensate. Using theoretical arguments supported by numerical simulation, we show that, the Kolmogorov-Zakharov solution is stable with respect to small isotropic perturbations in the absence of a condensate. Conversely, introducing a condensate component leads to a growth of the condensate occupation and ultimately destabilizes the inverse cascade. The system then evolves towards a new steady-state whose properties depend on the condensate dissipation rate.

nlin.CD↗
arXiv2026

Partial Multistep Prediction of Chaotic Time Series: What Remains Predictable beyond the Horizon of Predictability?

Sensitive dependence on initial conditions limits long-horizon prediction of chaotic systems even for deterministic dynamics, yet some dynamical information may remain recoverable after precise state information is lost. Recursive multistep methods are fragile because they reuse each intermediate prediction as input, allowing local errors to propagate through the rollout. We investigate whether partial multistep prediction can recover predictive information beyond the predictability horizon by omitting intermediate positions instead of assigning them unreliable values. The method uses a lightweight pattern-based local predictor. We analyze partial trajectories generated by different omission patterns and quantify the trade-off between accuracy on retained positions and the fraction classified as non-predictable. Experiments on the Lorenz system show that admissible masks contain trajectories that remain accurate far beyond the horizon of predictability, and an agreement-based ensemble recovers part of this information without future observations at the cost of a growing rejected fraction. At short horizons a broad accepted subset retains accuracy at moderate rejection, whereas at long horizons positive geometric skill requires increasingly strong selection. In this regime retained predictions no longer localize the state within a Lorenz scroll but still identify the occupied scroll, so topological structure outlasts geometric accuracy. A noise-based control reproduces this survival only after tuning of its perturbation scale, unlike masking. Across a heterogeneous collection of chaotic systems the ensemble strictly extends the positive-$R^2$ horizon of the recursive baseline for every retained system.

nlin.CD↗