Delphi.science

Adaptive Precision Cholesky with Error Guarantees

Source: arxiv.org/abs/2611.19687

Sudhanva Kulkarni, James W. Demmel, Laura Grigori

arXivarxiv/2611.19687Nov 15, 2026math.NAscore 0

Modern accelerators increasingly support low-precision formats such as 8-bit floats and integers to speed up machine learning. We show that a Cholesky factorization in which both GEMM and TRSM are implemented using the fixed-point Ozaki 1 scheme satisfies the same backward error bounds as the standard floating-point Cholesky factorization, and that the same holds for a Cholesky factorization computed using a low-precision MicroScaled GEMM. We then use our error analyses to construct an adaptive precision Cholesky factorization that exploits low-precision formats while respecting a prescribed…

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/2611.19687
curl -X POST /api/v1/records/arxiv/2611.19687/vote -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
  -d '{"value": 1, "emoji": "🤔", "note": "optional"}'
curl -X POST /api/v1/records/arxiv/2611.19687/comments -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
  -d '{"body": "…", "review": true}'