资讯

Threshold Compatibility and Support-Selection Diagnostics in Neural Discovery of Partial Differential Equations

Computer Science > Machine Learning

[Submitted on 27 Oct 2026]

Title:Threshold Compatibility and Support-Selection Diagnostics in Neural Discovery of Partial Differential Equations

View PDF HTML (experimental)Abstract:Accurate regression need not yield the correct equation after support selection. For a median/sample-standard-deviation threshold in neural PDE discovery, we derive an exact two-signal retention boundary and a complete-vector perturbation bound. Exact advection-diffusion libraries lose diffusion despite coefficient recovery to numerical precision. True-support refits of forty saved sparse states reduce diffusion error without restoring all derivative accuracy. A fixed sequential threshold ridge comparator recovers exact support in sixty ideal controls but selects extra terms in every learned state. All-fold correct-support candidates are rejected by conditional validation; post hoc energy pooling preserves their higher minimum loss in all sixty states. Supplied-support estimates outperform unchanged selected equations in all 240 physical-residual comparisons on four unused analytic trajectories. With coefficients, observation points and weights fixed, exact replacement of the learned library and target reverses their learned-row ordering in 59/60 states. Weighting both losses also fails the specified Burgers diffusion-benefit criterion at every tested budget. These finite diagnostics distinguish threshold incompatibility, refit damage and score rejection; they establish neither automatic correction nor general recovery or predictive stability.

Ancillary-file links:

Ancillary files (details):

- DEFINITIONS_AND_LIMITS.csv

- FORMAL_VERIFICATION.txt

- METHODS_AND_PROVENANCE.txt

- README.txt

- REFERENCE_SOURCE_RECORD.txt

- REPRODUCIBILITY_NOTES.txt

- TABLE_INDEX.csv

- literature_contribution_matrix.csv

- literature_diagnostics/PRIMARY_SOURCE_INSPECTION.txt

- literature_diagnostics/README.txt

- literature_diagnostics/SIGN_SEMANTIC_AUDIT.txt

- literature_diagnostics/SIGN_WITNESS.txt

- literature_diagnostics/audit.json

- literature_diagnostics/exact_sign_witness.json

- literature_diagnostics/input_hashes.json

- literature_diagnostics/pooled_candidate_scores.csv

- literature_diagnostics/pooled_state_scores.csv

- literature_diagnostics/summary.json

- score_outputs/anomalies.csv

- score_outputs/controls.csv

- score_outputs/crossed_scores.csv

- score_outputs/domain_score_summary.csv

- score_outputs/external_score_replay.csv

- score_outputs/old_score_replay.csv

- score_outputs/pairwise_group_summary.csv

- score_outputs/pairwise_rankings.csv

- score_outputs/same_point_scores.csv

- score_outputs/score_summary_long.csv

- score_outputs/six_group_summary.csv

- score_outputs/state_diagnostics.csv

- tables/S_AD_coefficients.md

- tables/S_AD_conditional.md

- tables/S_AD_derivatives.md

- tables/S_AD_precision.md

- tables/S_AD_projection.md

- tables/S_B_all_pairs.csv

- tables/S_B_all_runs.csv

- tables/S_B_coefficients.md

- tables/S_B_fields.md

- tables/S_B_pair_ratios.md

- tables/S_CD_all_pairs.csv

- tables/S_CD_descriptive_pooling.csv

- tables/S_CD_descriptive_pooling.md

- tables/S_CD_gates.md

- tables/S_CD_summary.csv

- tables/S_CD_summary.md

- tables/S_C_HT_coefficients.md

- tables/S_C_HT_fields.md

- tables/S_C_H_coefficients.md

- tables/S_C_H_fields.md

- tables/S_D_H_coefficients.md

- tables/S_D_H_fields.md

- tables/S_E_Q1_U_by_mask.csv

- tables/S_E_Q1_U_by_mask.md

- tables/S_E_amplification.md

- tables/S_E_archived_negative_cases.csv

- tables/S_E_diagnostic_reversals.csv

- tables/S_E_diagnostic_reversals.md

- tables/S_E_negative_cases.md

- tables/S_E_ordered_summary.csv

- tables/S_E_ordered_summary.md

- tables/S_F_K_candidates.csv

- tables/S_F_K_missing.md

- tables/S_F_all_methods.csv

- tables/S_F_block1_downstream.md

- tables/S_F_block1_upstream.md

- tables/S_F_block2_downstream.md

- tables/S_F_block2_upstream.md

- tables/S_F_guardrails.csv

- tables/S_F_guardrails.md

- tables/S_F_selection.csv

- tables/S_F_selection.md

- tables/S_F_selector_all_eight_ratios.csv

- tables/S_F_selector_all_eight_ratios.md

- tables/S_ORACLE_exact.md

- tables/S_ORACLE_phases.md

- tables/S_RB_coefficients.csv

- tables/S_RB_coefficients.md

- tables/S_RB_fields.csv

- tables/S_RB_fields.md

- tables/S_RB_moments.csv

- tables/S_RB_moments.md

- tables/S_SI_endpoints_1.csv

- tables/S_SI_endpoints_1.md

- tables/S_SI_endpoints_2.csv

- tables/S_SI_endpoints_2.md

- tables/S_SI_guards_1.csv

- tables/S_SI_guards_1.md

- tables/S_SI_guards_2.csv

- tables/S_SI_guards_2.md

- tables/S_SI_phase_1.csv

- tables/S_SI_phase_1.md

- tables/S_SI_phase_2.csv

- tables/S_SI_phase_2.md

- tables/S_WL_coefficients.csv

- tables/S_WL_coefficients.md

- tables/S_WL_comparisons.csv

- tables/S_WL_comparisons.md

- tables/S_WL_fields.csv

- tables/S_WL_fields.md

- tables/S_WL_guardrails.csv

- tables/S_WL_guardrails.md

- tables/S_WL_losses.csv

- tables/S_WL_losses.md

- tables/S_external_states_1.csv

- tables/S_external_states_1.md

- tables/S_external_states_2.csv

- tables/S_external_states_2.md

- tables/S_external_states_3.csv

- tables/S_external_states_3.md

- tables/S_external_trajectories.csv

- tables/S_external_trajectories.md

- tables/S_new_budget_all.md

- tables/S_new_budget_guardrails.md

- tables/S_new_budget_secondary.md

- tables/S_new_budget_transitions.md

- tables/S_new_conditional_gaps.md

- tables/S_same_point_summary.md

- tables/S_score_gap.csv

- tables/S_score_gap.md

- tables/S_selector_cases_1.csv

- tables/S_selector_cases_1.md

- tables/S_selector_cases_2.csv

- tables/S_selector_cases_2.md

- tables/S_selector_cases_3.csv

- tables/S_selector_cases_3.md

- tables/S_selector_controls.csv

- tables/S_selector_controls.md

- tables/S_selector_sensitivity.csv

- tables/S_selector_sensitivity.md

- tables/T1_design.csv

- tables/T1_design.md

- tables/T2_RB_cells.csv

- tables/T2_RB_cells.md

- tables/T3_WL_summary.csv

- tables/T3_WL_summary.md

- tables/T4_WL_gates.csv

- tables/T4_WL_gates.md

- tables/T5_budget.md

- tables/T6_conditional.md

- tables/T7_analytic.md

- tables/T8_analytic_conditional.md

- tables/T_concise_evidence.md

- tables/T_design_hierarchy.csv

- tables/T_design_hierarchy.md

- tables/T_external_summary.csv

- tables/T_external_summary.md

- tables/T_intervention_blocks.csv

- tables/T_intervention_blocks.md

- tables/T_selector_blocks.csv

- tables/T_selector_blocks.md

- tables/external_states.csv

- tables/external_trajectory_scores.csv

- tables/selector_score_gaps.csv

Additional Features

References & Citations

Loading...

Bibliographic and Citation Tools

Bibliographic Explorer (What is the Explorer?)

Connected Papers (What is Connected Papers?)

Litmaps (What is Litmaps?)

scite Smart Citations (What are Smart Citations?)

Code, Data and Media Associated with this Article

alphaXiv (What is alphaXiv?)

CatalyzeX Code Finder for Papers (What is CatalyzeX?)

DagsHub (What is DagsHub?)

Gotit.pub (What is GotitPub?)

Hugging Face (What is Huggingface?)

ScienceCast (What is ScienceCast?)

Demos

Recommenders and Search Tools

Influence Flower (What are Influence Flowers?)

CORE Recommender (What is CORE?)

IArxiv Recommender

(What is IArxiv?)

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.