Metraipolis › archive › dc_f65660f2be50
Quantization Loss in Safety-Critical Decision Systems (2026) · DOI 10.31224/8259
- Source
- Metraipolis archive
- Original
- https://doi.org/10.31224/8259
- License (as archived)
- CC-BY-4.0
- Archived
- 2026-10-08T01:16:01.502Z
- SHA-256 of the archived text
36c94db3ac16ca2bfbd2cfd35ab59cf881645049b82fc93274aaeea26ff96791- Stable URL of this page
https://metraipolis.com/documento/dc_f65660f2be50
How to cite
Quantization Loss in Safety-Critical Decision Systems (2026) · DOI 10.31224/8259 — Metraipolis archive, https://doi.org/10.31224/8259 (CC-BY-4.0); archived copy: https://metraipolis.com/documento/dc_f65660f2be50 (Metraipolis archive dc_f65660f2be50, sha256 36c94db3ac16ca2bfbd2cfd35ab59cf881645049b82fc93274aaeea26ff96791)
BibTeX
@misc{dcf65660f2be50,
title = {Quantization Loss in Safety-Critical Decision Systems (2026) · DOI 10.31224/8259},
howpublished = {Metraipolis archive},
url = {https://doi.org/10.31224/8259},
note = {License CC-BY-4.0; archived copy https://metraipolis.com/documento/dc_f65660f2be50; sha256 36c94db3ac16ca2bfbd2cfd35ab59cf881645049b82fc93274aaeea26ff96791},
year = {2026}
}Archived text
Reproduced under CC-BY-4.0. Attribution: Metraipolis archive, https://doi.org/10.31224/8259. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Articolo recente con licenza aperta: scheda dai metadati Crossref (posted-content). Titolo: Quantization Loss in Safety-Critical Decision Systems Autori: Greg Passmore Pubblicato: 2026-09-18 · editore: Open Engineering Inc DOI: https://doi.org/10.31224/8259 Licenza dell'articolo: CC-BY-4.0 Abstract (testo degli autori, CC-BY-4.0): Binary decision aids, that is, indicators that reduce a graded measurement to a category or a bit before an operator sees it, appear across safety-critical engineering. Radiation-therapy interlocks, reactor status lights, cockpit stall warnings, and combat identification systems all share this pattern. They also share a failure mode near the threshold, that is, the regime where the reduction discards the information a rational decision-maker needs to act. This paper treats the problem as one of information loss under quantization, examines four well-documented incidents as illustrative evidence, and argues for a design rule, that is, where a graded measurement is available, carry the confidence forward and present it alongside the decision rather than only the argmax. The four incidents help illustrate the argument; they do not attempt to prove it in a statistical sense. Fonte: https://api.crossref.org/works/10.31224%2F8259 — metadati Crossref, consultati il 2026-10-07. Interrogazione: «efficient inference quantization».