-
Notifications
You must be signed in to change notification settings - Fork 0
/
README.ibexopt
120 lines (96 loc) · 5.09 KB
/
README.ibexopt
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
IbexOpt is a solver based on IBEX versions 2.8.9 that can be used
either "stand-alone" or with AMPL's options to solve nonlinear
problems expressed in AMPL.
To use IbexOpt with AMPL, you have several options. You can invoke
it within an AMPL session by saying
solve;
or, if $solver is not already ibexopt,
option solver ibexopt;
solve;
When $solver has its default value (ibexopt), AMPL's solve command invokes
ibexopt stub -AMPL
If a stub is present, IbexOpt tries to write the computed solution to
stub.sol
For details not given here, see the IBEX documentation:
http://www.ibex-lib.org
Source for IBEX is available from
https://github.com/ibex-team/
-----------------------
solve_result_num values
=======================
Here is a table of solve_result_num values that "ibexopt" can return
to an AMPL session, along with the text that appears in the associated
solve_message.
Value Message
0 OPTIMIZATION SUCCESS!
The global minimum (with respect to the precision
required) has been found. In particular, at least
one feasible point has been found, less
than obj_init_bound, and in the time limit.
200 INFEASIBLE PROBLEM.
No feasible point exist less than initial_loup.
In particular, the function returns INFEASIBLE
if the initial bound \"obj_init_bound\" is LESS
than the true minimum (this case is only possible
if abs_eps_f and rel_eps_f are 0). In the
latter case, there may exist feasible points.
201 NO FEASIBLE POINT FOUND.
No feasible point could be found less than
obj_init_bound. Contrary to INFEASIBLE,
infeasibility is not proven here. Warning: this
return value is sensitive to the abs_eps_f and
rel_eps_f parameters. The upperbounding makes the
optimizer only looking for points less than
min{(1-rel_eps_f)*obj_init_bound,obj_init_bound-abs_eps_f}.
300 UNBOUNDED OBJECTIVE FONCTION.
The objective function seems unbounded (tends to -oo).
400 TIMEOUT.
time limit is reached.
402 UNREACHED PRECISION.
The search is over but the resulting interval [uplo,loup]
does not satisfy the precision requirements. There are
several possible reasons: the goal function may be too
pessimistic or the constraints function may be too
pessimistic with respect to the precision requirement
(which can be too stringent). This results in tiny boxes
that can neither be contracted nor used as new loup candidates.
Finally, the eps_x parameter may be too large.
-----------------------
ibexopt option
=======================
To set these directives, assign a string specifying their values to the AMPL option ibexopt_options. For example:
option ibexopt_options 'rel_eps_f=1e-5 timeout=1000 rigor=1 simpl_level=2';
Keywords are followed by a value except those marked No value is expected in the listing.
rel_eps_f Relative precision on the objective. Default: 1e-3.
abs_eps_f Absolute precision on the objective function. Default: 1.e-7.
eps_h Relaxation value of the equality constraints. Default: 1.e-8.
random_seed Random seed (useful for reproducibility). Default: 1.
timeout Timeout (time in seconds). Default: -1 (none).
obj_numb Choose which objective function of the AMPL model: 0 = none, 1 = first. Default: 1.
simpl_level Expression simplification level. Possible values are:
0 : no simplification at all (fast).
1 : basic simplifications (fairly fast). E.g. x+1+1 --> x+2
2 : more advanced simplifications without developing (can be slow).
E.g. x*x + x^2 --> 2x^2
3 : simplifications with full polynomial developing (can blow up!).
E.g. x*(x-1) + x --> x^2
Default value is : 1.
init_obj_value Initialization of the upper bound of the objectif function
with a known value. Default: +infinity.
rigor Activate rigor mode (certify feasibility of equalities). Default: -1.
-1 : By default: #default_rigor.
0 : Deactivate rigor mode
1 : Activate rigor mode, feasibility of equalities is certified.
kkt Activate contractor based on Kuhn-Tucker conditions. Default: -1.
-1 : By default, we apply KKT only for unconstrained problems.
0 : deactivate KKT contractor
1 : activate KKT contractor
inHC4 Activate feasibility search with LoupFinderInHC4. Default: -1.
-1 : By default: #default_inHC4.
0 : Deactivate inHC4
1 : Activate inHC4, feasibility is also tried with LoupFinderInHC4.
trace Activate trace. Updates of lower and upper bound are printed while minimizing. Default: 1.
0 : nothing is printed
1 : prints every loup/uplo update.
2 : prints also each handled node (warning: can generate very long trace).
-------------------------------