Causal Contribution and Transfer-Entropy-Inspired Diagnostics for Inverse Nonlinear Schrödinger Problems

Inverse problems for nonlinear partial differential equations often rely on time-resolved measurements, yet the informational value of those measurements is rarely uniform across time. In this work, we develop a causal-contribution framework, inspired by transfer entropy, for quantifying how much each newly assimilated snapshot improves recovery of the unknown potential in a nonlinear Schrödinger equation (NLSE). Our inverse solver is an operator-guided neural architecture, NLS-Net, whose layers mirror a Strang-splitting discretization: linear dispersive propagation is represented by convolutional blocks, while the cubic nonlinearity and external potential are handled through pointwise phase updates. The unknown potential is parameterized as a sparse expansion over a prescribed library of trigonometric and localized Gaussian atoms, reducing functional inversion to coefficient identification with proximal ℓ 1 {\ell^{1}} regularization. Within this setting, increasing the amount of input data can mean either enriching the measurement set or increasing the fidelity of the operator-guided forward block. We show that the first type of increase does not translate into uniform information gain: some snapshots yield large conditional improvement, whereas others are nearly redundant after conditioning on earlier data. Building on this observation, we introduce a solver-agnostic and a contribution-based snapshot selection rule to examine the tradeoff between observation budget and reconstruction accuracy. Numerical experiments on inverse NLSE problems reveal strongly non-uniform temporal contributions and show that the effectiveness of reduced snapshot sets depends on the observation budget. The connection to transfer entropy lies in conditioning on previously assimilated observations, rather than in estimating an entropy-based quantity.

Authors

Institutions

Publication Details

Journal
Computational Methods in Applied Mathematics
Published
2026-11-18
DOI
https://doi.org/10.1515/cmam-2026-0166
Primary Topic
Numerical methods in inverse problems
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Causal Contribution and Transfer-Entropy-Inspired Diagnostics for Inverse Nonlinear Schrödinger Problems

Y. H. Wang, Eric T. Chung, Siqing Liu
2 links
Computational Methods in Applied Mathematics
Numerical methods in inverse problems
article

Causal Contribution and Transfer-Entropy-Inspired Diagnostics for Inverse Nonlinear Schrödinger Problems

Y. H. Wang, Eric T. Chung, Siqing Liu
article en

Abstract

Inverse problems for nonlinear partial differential equations often rely on time-resolved measurements, yet the informational value of those measurements is rarely uniform across time. In this work, we develop a causal-contribution framework, inspired by transfer entropy, for quantifying how much each newly assimilated snapshot improves recovery of the unknown potential in a nonlinear Schrödinger equation (NLSE). Our inverse solver is an operator-guided neural architecture, NLS-Net, whose layers mirror a Strang-splitting discretization: linear dispersive propagation is represented by convolutional blocks, while the cubic nonlinearity and external potential are handled through pointwise phase updates. The unknown potential is parameterized as a sparse expansion over a prescribed library of trigonometric and localized Gaussian atoms, reducing functional inversion to coefficient identification with proximal ℓ 1 {\ell^{1}} regularization. Within this setting, increasing the amount of input data can mean either enriching the measurement set or increasing the fidelity of the operator-guided forward block. We show that the first type of increase does not translate into uniform information gain: some snapshots yield large conditional improvement, whereas others are nearly redundant after conditioning on earlier data. Building on this observation, we introduce a solver-agnostic and a contribution-based snapshot selection rule to examine the tradeoff between observation budget and reconstruction accuracy. Numerical experiments on inverse NLSE problems reveal strongly non-uniform temporal contributions and show that the effectiveness of reduced snapshot sets depends on the observation budget. The connection to transfer entropy lies in conditioning on previously assimilated observations, rather than in estimating an entropy-based quantity.

Computational Methods in Applied Mathematics
University of Alabama (US), Chinese University of Hong Kong (HK)
Numerical methods in inverse problems
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon