1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
|
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML3.2 EN">
<HTML>
<HEAD> <link rel="canonical" href="http://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/KSP/KSPQCG.html" />
<META NAME="GENERATOR" CONTENT="DOCTEXT">
<TITLE>KSPQCG</TITLE>
</HEAD>
<BODY BGCOLOR="FFFFFF">
<div id="version" align=right><b>petsc-3.7.5 2017-01-01</b></div>
<div id="bugreport" align=right><a href="mailto:petsc-maint@mcs.anl.gov?subject=Typo or Error in Documentation &body=Please describe the typo or error in the documentation: petsc-3.7.5 v3.7.5 docs/manualpages/KSP/KSPQCG.html "><small>Report Typos and Errors</small></a></div>
<A NAME="KSPQCG"><H1>KSPQCG</H1></A>
Code to run conjugate gradient method subject to a constraint on the solution norm. This is used in Trust Region methods for nonlinear equations, <A HREF="../SNES/SNESNEWTONTR.html#SNESNEWTONTR">SNESNEWTONTR</A>
<H3><FONT COLOR="#CC3333">Options Database Keys</FONT></H3>
<DT><B>-ksp_qcg_trustregionradius <r> </B> -Trust Region Radius
<br>
<P>
Notes: This is rarely used directly
<P>
<P>
Notes: Use preconditioned conjugate gradient to compute
an approximate minimizer of the quadratic function
<P>
q(s) = g^T * s + .5 * s^T * H * s
<P>
subject to the Euclidean norm trust region constraint
<P>
|| D * s || <= delta,
<P>
where
<P>
delta is the trust region radius,
g is the gradient vector, and
H is Hessian matrix,
D is a scaling matrix.
<P>
<A HREF="../KSP/KSPConvergedReason.html#KSPConvergedReason">KSPConvergedReason</A> may be
<pre>
KSP_CONVERGED_CG_NEG_CURVE if convergence is reached along a negative curvature direction,
</pre>
<pre>
KSP_CONVERGED_CG_CONSTRAINED if convergence is reached along a constrained step,
</pre>
<pre>
other <A HREF="../KSP/KSP.html#KSP">KSP</A> converged/diverged reasons
</pre>
<P>
<P>
<H3><FONT COLOR="#CC3333">Notes</FONT></H3>
<H3><FONT COLOR="#CC3333">Currently we allow symmetric preconditioning with the following scaling matrices</FONT></H3>
<A HREF="../PC/PCNONE.html#PCNONE">PCNONE</A>: D = Identity matrix
<A HREF="../PC/PCJACOBI.html#PCJACOBI">PCJACOBI</A>: D = diag [d_1, d_2, ...., d_n], where d_i = sqrt(H[i,i])
<A HREF="../PC/PCICC.html#PCICC">PCICC</A>: D = L^T, implemented with forward and backward solves.
Here L is an incomplete Cholesky factor of H.
<P>
<H3><FONT COLOR="#CC3333">References</FONT></H3>
<DT><B>1. </B> -Trond Steihaug, The Conjugate Gradient Method and Trust Regions in Large Scale Optimization,
SIAM Journal on Numerical Analysis, Vol. 20, No. 3 (Jun., 1983).
<br>
<P>
<H3><FONT COLOR="#CC3333">See Also</FONT></H3>
<A HREF="../KSP/KSPCreate.html#KSPCreate">KSPCreate</A>(), <A HREF="../KSP/KSPSetType.html#KSPSetType">KSPSetType</A>(), <A HREF="../KSP/KSPType.html#KSPType">KSPType</A> (for list of available types), <A HREF="../KSP/KSP.html#KSP">KSP</A>, <A HREF="../KSP/KSPQCGSetTrustRegionRadius.html#KSPQCGSetTrustRegionRadius">KSPQCGSetTrustRegionRadius</A>()
<BR><A HREF="../KSP/KSPQCGGetTrialStepNorm.html#KSPQCGGetTrialStepNorm">KSPQCGGetTrialStepNorm</A>(), <A HREF="../KSP/KSPQCGGetQuadratic.html#KSPQCGGetQuadratic">KSPQCGGetQuadratic</A>()
<P><B><P><B><FONT COLOR="#CC3333">Level:</FONT></B>developer
<BR><FONT COLOR="#CC3333">Location:</FONT></B><A HREF="../../../src/ksp/ksp/impls/qcg/qcg.c.html#KSPQCG">src/ksp/ksp/impls/qcg/qcg.c</A>
<BR><A HREF="./index.html">Index of all KSP routines</A>
<BR><A HREF="../../index.html">Table of Contents for all manual pages</A>
<BR><A HREF="../singleindex.html">Index of all manual pages</A>
</BODY></HTML>
|
