-
Notifications
You must be signed in to change notification settings - Fork 0
/
get_par.m
137 lines (127 loc) · 6.04 KB
/
get_par.m
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
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
function par = get_par(seednum,probnum,n,maxit)
%------------------------------
% problem description
%------------------------------
par.problem=probnum;
% 3: standard quadratic function with diagonal matrix A and
% paraboloid coordinate transformation (problem B from the
% paper, arXiv:1106.4426v2)
par.probPars{1}=[n 0.0]; % problemA
par.probPars{2}=[n 1.0 0.0]; % problemB
par.probPars{3}=[n 0.0]; % problemC
par.probPars{4}=[n 0.0]; % problemD
par.probPars{5}=[n 0.0]; % problemE
par.probPars{6}=[n 0.0]; % problemF
par.probPars{7}=[n nan]; % problemG
par.probPars{8}=[n 0.0]; % problemQuadratic
par.probNames = {'A', 'B', 'C', 'D', 'E', 'F', 'G', 'Q'};
%------------------------------
% parameters
%------------------------------
% some general parameters
% first select which methods to compare
par.compareNGMRES_sdls=1; % run 1
par.compareNGMRES_sd=1; % run 2
par.compareNCG=1; % run 4
par.compare_sdls=0; % run 5
par.compareLBFGS=1; % run 7
par.compareNGMRESO_sdls=1; % run 8
par.compareNGMRESO_sd=1; % run 9
% running parameters for the Tensor test problem
par.compareNGMRESO_ALS=0; % run 10
par.compareNGMRES_ALS=0; % run 11
par.compareALS=0; % run 12
par.initSeed = seednum;
par.figStart=100; % figure number
par.precStep1=1e-4; % step in steepest descent preconditioner without line search
par.precStep2=1e0; % step factor in steepest descent preconditioner without line search
par.maxfev=20; % maximum function evaluations in all line searches
par.restartfigure=false; % indicate restarts on the pictures
par.epsi=1e-16; % for plotting convergence to f*
% line styles
par.fs{1}=':k';
par.fs{2}='-k';
par.fs{4}='-.k';
par.fs{5}='--k';
par.fs{7}='-vk';
par.fs{8}=':ok';
par.fs{9}='-ok';
par.fs{10}=':k'; % overlaps with 1!
par.fs{11}='--k'; % overlaps with 5!
par.fs{12}=':ok'; % overlaps with 8!
par.mispace{1} = 10;
par.mispace{2} = 10;
par.mispace{4} = 10;
par.mispace{5} = 10;
par.mispace{7} = 10;
par.mispace{8} = 10;
par.mispace{9} = 10;
par.mispace{10} = 10;
par.mispace{11} = 10;
par.mispace{12} = 10;
% ngmres parameters
par.par_ngmres.epsi=1e-12; % Regularization parameter for normal system
par.par_ngmres.w=20; % maximum window size
par.par_ngmres.maxIt=maxit; % maximum number of iterations
par.par_ngmres.relfTol=-1; % stopping tolerance on relative change in f
par.par_ngmres.absgTol=1e-14; % stopping tolerance on the
% 2-norm of g (scaled by 1/n)
par.par_ngmres.verbose=0; % level of output to screen (0=no output, 1=summary output,
% 2=detailed output for each iteration)
par.par_ngmres.logfev=true; % flag to include the history of the number of f and g
% evaluations in the output (true or false)
par.par_ngmres.logf=true; % flag to include the history of functiona values in the output
% (true or false)
par.par_ngmres.logg=true; % flag to include the history of the 2-norm of the gradient
par.par_ngmres.logRestart=true; % flag to include the history of the restarts
% ngmres line search parameters (for poblano_linesearch)
par.par_ngmres.LineSearch_ftol=1.000000000000000e-04;
par.par_ngmres.LineSearch_gtol=1e-1; % De Sterck paper default 1e-2
par.par_ngmres.LineSearch_initialstep=1;
par.par_ngmres.LineSearch_maxfev=par.maxfev;
par.par_ngmres.LineSearch_method='more-thuente';
par.par_ngmres.LineSearch_stpmax=1.000000000000000e+15;
par.par_ngmres.LineSearch_stpmin=1.000000000000000e-15;
par.par_ngmres.LineSearch_xtol=1.000000000000000e-15;
% ncg parameters (for comparing ncg with ngmres)
par.par_ncg=ncg('defaults');
par.par_ncg.MaxFuncEvals=50000;
par.par_ncg.MaxIters=par.par_ngmres.maxIt;
par.par_ncg.TraceFuncEvals=true;
par.par_ncg.TraceFunc=true;
par.par_ncg.TraceGradNorm=true;
par.par_ncg.RelFuncTol=par.par_ngmres.relfTol;
par.par_ncg.StopTol=par.par_ngmres.absgTol;
par.par_ncg.Display='off'; % iter, final or off
par.par_ncg.RestartNW=false;
par.par_ncg.Update='PR'; % 'PR', 'FR', 'HS' or 'SD'
par.par_ncg.LineSearch_maxfev=par.maxfev; % default: 20
par.par_ncg.LineSearch_ftol=1e-4; % default: 1e-4
par.par_ncg.LineSearch_gtol=1e-1; % De Sterck paper default 1e-2
par.par_ncg.LineSearch_initialstep=1;
% lbfgs parameters (for comparing lbfgs with ngmres)
par.par_lbfgs=lbfgs('defaults');
par.par_lbfgs.M=5; % window size
par.par_lbfgs.MaxFuncEvals=50000;
par.par_lbfgs.MaxIters=par.par_ngmres.maxIt;
par.par_lbfgs.TraceFuncEvals=true;
par.par_lbfgs.TraceFunc=true;
par.par_lbfgs.TraceGradNorm=true;
par.par_lbfgs.RelFuncTol=par.par_ngmres.relfTol;
par.par_lbfgs.StopTol=par.par_ngmres.absgTol;
par.par_lbfgs.Display='off'; % iter, final or off
par.par_lbfgs.LineSearch_maxfev=par.maxfev; % default: 20
par.par_lbfgs.LineSearch_ftol=1e-4; % default: 1e-4
par.par_lbfgs.LineSearch_gtol=1e-1; % De Sterck paper default 1e-2
par.par_lbfgs.LineSearch_initialstep=1;
% steepest decent line search parameters (for using SD as a preconditioner)
par.par_sdls.LineSearch_ftol=1.000000000000000e-04;
par.par_sdls.LineSearch_gtol=1e-1; % De Sterck paper default 1e-2
par.par_sdls.LineSearch_initialstep=1;
par.par_sdls.LineSearch_maxfev=par.maxfev;
par.par_sdls.LineSearch_method='more-thuente';
par.par_sdls.LineSearch_stpmax=1.000000000000000e+15;
par.par_sdls.LineSearch_stpmin=1.000000000000000e-15;
par.par_sdls.LineSearch_xtol=1.000000000000000e-15;
par.par_als.maxIt = par.par_ngmres.maxIt;
end