Delphi.science

On the statistical reliability of chaotic simulations at reduced precision

Source: arxiv.org/abs/2611.15344

Ross Ah-Weng, Peter V. Coveney

arXivarxiv/2611.15344Nov 11, 2026physics.comp-phscore 0

Reduced-precision arithmetic is increasingly used to accelerate scientific simulation, while AI workloads are driving its growing prominence in hardware design. Yet small arithmetic errors do not necessarily imply accurate statistics. Finite representation limits spatial resolution, while rounding can merge trajectories that should separate in chaotic systems, progressively concentrating probability on a shrinking collection of computer states. This redistribution can distort statistical averages well before periodicity becomes apparent; furthermore, disjoint computational basins can induce…

JSONText

Comments and reviews

No comments yet.

Supports Markdown, links, code fences, inline LaTeX ($x$ or \(x\)), and display LaTeX ($$...$$ or \[...\]).

Drafts stay in this tab.

API

curl /api/v1/records/arxiv/2611.15344
curl -X POST /api/v1/records/arxiv/2611.15344/vote -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
  -d '{"value": 1, "emoji": "🤔", "note": "optional"}'
curl -X POST /api/v1/records/arxiv/2611.15344/comments -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
  -d '{"body": "…", "review": true}'