From: "Bruno Luong" <b.luong@fogale.findmycountry>
Newsgroups: comp.soft-sys.matlab
Subject: Re: Non-linear optimization
Date: Thu, 7 Mar 2013 22:52:07 +0000 (UTC)
Organization: FOGALE nanotech
Lines: 10
Message-ID: <khb5mn$7bg$>
References: <kh2m44$4eh$> <kh32ds$hrn$> <kh46mg$s0t$> <kh89he$4al$> <kh8bsc$bov$> <kh8dvs$imu$> <kh8g3b$p6i$> <kh8hi7$t86$> <kh8jcc$4qj$> <kh8skk$s6c$> <kh9it4$qg1$> <khaui3$efe$> <khavr6$imm$> <khb188$n3q$> <khb24e$q0a$> <khb4uc$55k$>
Reply-To: "Bruno Luong" <b.luong@fogale.findmycountry>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
X-Trace: 1362696727 7536 (7 Mar 2013 22:52:07 GMT)
NNTP-Posting-Date: Thu, 7 Mar 2013 22:52:07 +0000 (UTC)
X-Newsreader: MATLAB Central Newsreader 390839
Xref: comp.soft-sys.matlab:790603

"Matt J" wrote in message <khb4uc$55k$>...

> I don't see how that can be. For non-convex functions and non-posdef Hessians, an empirically chosen lambda could easily leave H+lambda*I singular, or at least, not positive definite and therefore non-descending. You would have to choose lambda>=min(eig(H)) to be sure that didn't happen, and that would require an eigen-analysis of H.

The strategy and rules to chose lambda is given by LM algorithm. You can check the textbook that explains all the details.

Again all that is well known.