conjugate gradient convergence

*To*: mathgroup at smc.vnet.net*Subject*: [mg54036] conjugate gradient convergence*From*: knight_aziz at hotmail.com (knighty)*Date*: Tue, 8 Feb 2005 05:31:19 -0500 (EST)*Sender*: owner-wri-mathgroup at wolfram.com

hello! I implemented the CG algorithm (with no preconditionning) for test purposes, just to know how it works. i tried it with symetric but not definite positive matrices. it SEEMS to work well, althought it is established that it converges for symetric definite positive matrices (Only?). Since it's a quite old algorithm, i gess this fact was already noticed. So the question is: Why it's sayed that definite positivenes (i'm not sure for the speling :/) is necessary for CG to converge ?