Information-Complete Computational Discovery
Year: 2026
Type: preprint
Abstract: Scientific discovery depends on retaining the distinctions that a model has not yet explained. We formulate the Information-Complete Computational Discovery Engine (ICCDE), a general methodology reconstructed from the K-Field corpus and expressed as an auditable sequence of structural certification, matched contrasts, frozen transfer, residual decomposition, inverse inference, and probe design. Information completeness is defined operationally as recoverability of the retained measurement record and its transformations. Five layers separate registry, metric, dynamics, readout, and inference. A linear residual-to-probe result distinguishes discrepancies that parameter changes can absorb from consistency coordinates that they cannot alter, and gives the exact number of unresolved directions removed by an added probe. Independent finite calculations reproduce the 27-node Cell3 registry, 49 endpoint relations, a 30-dimensional cycle space, and a four-dimensional affine midpoint kernel. Thirteen reciprocal measurements resolve six metric parameters while retaining seven consistency dimensions. A single-family perturbation produces an exact consistency norm squared of 4/525 in the declared units. Spectral moments recover the weighted relative inverse-gap variance 48/169 in a two-gap example. Schur reduction preserves hidden-state normalization, while a dynamic readout example shows how an instantaneously invisible state becomes observable. Twenty-seven computational checks accompany proofs, counterexamples, signed residual fields, and executable reconstruction code. The contribution is a reproducible discovery method that turns unresolved structure into constrained experimental questions, with explicit boundaries between mathematical identity, computational demonstration, and physical mechanism.A K-Field Method for Exact Structure, Residual Inference, and Probe Design. This standalone methods and computational discovery preprint contains 24 pages, 29 equations, eight tables, three figures, and 27 reproduced computational checks. The JSON retains source identities, complete small residual fields, audit corrections, and executable reconstruction code. The original K-Field sources are referenced and are not included.Licensing: CC BY 4.0 for the paper, abstract and non-code research content. Embedded program source in /reproduction_code, /figure_generation_code and /layout_generation_code is AGPL-3.0-only. Full license texts and component scope are in the JSON companion. Copyright 2026 Krytronx, LLC. Attribution: P. Dwayne Esterline, Krytronx, LLC. Contact: [email protected] release=faceaf29-b881-5f16-895c-d6ccfbcf83b1 plan=3ca0df30367479dcf3e550b6657f8b9c220c10bbe89ac85068b28de2421bbb8f payload=da21645ef3f8be611b33973e8a32851aae86328abf0cfead45f964c5ba9ca8afgithub: https://github.com/KrytronDwayne/Information-Complete-Computational-Discovery/releases/tag/v1.0.0Preferred citation: Esterline, P. Dwayne. Information-Complete Computational Discovery. Version 1.0.0. https://doi.org/10.5281/zenodo.24381256
Author: Esterline, P. Dwayne
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7236640499
