[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Octave-bug-tracker] [bug #45146] chol succeeds with incorrect factoriza
From: |
anonymous |
Subject: |
[Octave-bug-tracker] [bug #45146] chol succeeds with incorrect factorization on non-hermitian complex inputs |
Date: |
Tue, 19 May 2015 23:15:29 +0000 |
User-agent: |
Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Ubuntu Chromium/41.0.2272.76 Chrome/41.0.2272.76 Safari/537.36 |
URL:
<http://savannah.gnu.org/bugs/?45146>
Summary: chol succeeds with incorrect factorization on
non-hermitian complex inputs
Project: GNU Octave
Submitted by: None
Submitted on: Tue 19 May 2015 11:15:27 PM UTC
Category: Documentation
Severity: 3 - Normal
Priority: 5 - Normal
Item Group: Incorrect Documentation
Status: None
Assigned to: None
Originator Name: Nathan Whitehead
Originator Email: address@hidden
Open/Closed: Open
Discussion Lock: Any
Release: 3.8.1
Operating System: GNU/Linux
_______________________________________________________
Details:
The chol() function silently succeeds with non-intuitive results in the case
where the input matrix is complex and non-hermitian.
The documentation states that the function fails if the input is not positive
definite, or that chol(A)' * chol(A) = A if it is positive definite. For the
case where A is complex but not hermitian, the function succeeds and returns a
value that does not satisfy chol(A)' * chol(A) = A.
A=[10,-i;-i,10;]; norm(A - chol(A)'*chol(A), Inf)
For this code I would expect either an error or a result that is close to
machine epsilon. I actually get the result:
ans = 2.0000
I believe the function is only looking at the upper triangular part and
assuming the lower triangular part is conjugate transpose; this should be
included in the documentation for the function.
_______________________________________________________
Reply to this item at:
<http://savannah.gnu.org/bugs/?45146>
_______________________________________________
Message sent via/by Savannah
http://savannah.gnu.org/
- [Octave-bug-tracker] [bug #45146] chol succeeds with incorrect factorization on non-hermitian complex inputs,
anonymous <=