From: LU Urquhart on
Hello,

When is a condition number too large. From reading past posts, I found that one reader recommended, " "Large" is a word that here only has meaning in the context of your data and the numerical precision that you are working in. A "too large" condition number is usually considered to be roughly 1/eps."

Quoted from, "Subject: A question about condition number, From: John D'Errico, Date: 31 Oct, 2007 10:47:56, Message: 13 of 14"



In my case my data has probably 4 significant digits, I am working in double precision and my condition numbers are in the range of 1e9.

Most numerical method books I have found only use the word "large" to describe an ill conditioned matrix. Does anyone know of a reference which has more discussion on this topic.

Thank you!