About this resource
<p>Code and raw per-run records accompanying the manuscript <em>The Level Is the Choice: Reconstruction and Identification Prefer Opposite Discrepancy Levels in Three Misspecified Inverse Physics-Informed Neural Networks</em>.</p> <p>Contents:</p> <ul> <li><code>sandbox/</code> — problem definitions, the vectorised ensemble trainer, one driver per experiment (w1–w20), and the figure scripts</li> <li><code>results/</code> — raw per-run outputs, one JSON per experiment: every seed, every logged step</li> <li><code>scripts/</code> — the emitters that generate every numeric macro in the manuscript from those records, and the theory verification script</li> </ul> <p>All data are synthetic and regenerate deterministically from this code. Seeds fix the collocation and sensor point sets, the noise realisation and the network initialisation, so a seed identifies one run across every arm — which is what makes the reported statistics paired.</p> <p>Code MIT; recorded outputs CC BY 4.0.</p>
Files and access
Resources are read online through DARE. Direct file links are not published; access is checked against DSpace when a document is opened. Authorised librarian workflows are unaffected.
This record contains bibliographic metadata, but no files are currently attached.
