Status solved in sdpts but Infeasible in sedumi


(ly) #1

Using different parsers to run the same code has different results. Is there any problem with the code?

Successive approximation method to be employed.
For improved efficiency, SDPT3 is solving the dual problem.
SDPT3 will be called several times to refine the solution.
Original size: 3199 variables, 1391 equality constraints
200 exponentials add 1600 variables, 1000 equality constraints

Cones | Errors |
Mov/Act | Centering Exp cone Poly cone | Status
--------±--------------------------------±--------
0/ 0 | 0.000e+00 0.000e+00 0.000e+00 | Solved

Status: Solved
Optimal value (cvx_optval): +0.0160707

Successive approximation method to be employed.
For improved efficiency, SeDuMi is solving the dual problem.
SeDuMi will be called several times to refine the solution.
Original size: 3199 variables, 1391 equality constraints
200 exponentials add 1600 variables, 1000 equality constraints

Cones | Errors |
Mov/Act | Centering Exp cone Poly cone | Status
--------±--------------------------------±--------
0/ 0 | 0.000e+00 0.000e+00 0.000e+00 | Unbounded

Status: Infeasible
Optimal value (cvx_optval): +Inf


(Mark L. Stone) #2

Try Mosek if you can. Also try CVXQUAD which you have already used.


(ly) #3

Thanks for your reply ,I try it ,but it doesn’t work
Successive approximation method to be employed.
For improved efficiency, Mosek is solving the dual problem.
Mosek will be called several times to refine the solution.
Original size: 3199 variables, 1391 equality constraints
200 exponentials add 1600 variables, 1000 equality constraints

Cones | Errors |
Mov/Act | Centering Exp cone Poly cone | Status
--------±--------------------------------±--------
0/ 0 | 0.000e+00 0.000e+00 0.000e+00 | Unbounded

Status: Infeasible
Optimal value (cvx_optval): +Inf


(Mark L. Stone) #4

Well, maybe it is infeasible. You could try removing the objective function and see whether it is still reported infeasible. Also, try to invoke CVXQUAD with MOSEK and see what happens.


(ly) #5

Hello ,stone
when I removing the objection function (minimize 1)
I got the same result, the status is sloved with sdpt3 ,cvx_optval =1
but the sedumi and mosek infeasible
and I invoke CVXQUAD with mosek ,but there is no log function in my code


(Mark L. Stone) #6

You’ll need to show your code in order for me to determine what you need to do to get CVXQUAD’'s Pade approximation to be invoked instead of CVX’s successive approximation method…


(ly) #7

OK, I sent my code to your mailbox.


(Mark L. Stone) #8

No. Post here. Those are the forum rules.


(Mark L. Stone) #10

It looks like the only thing causing the successive approximation method to be invoked is exp of a CVX expression. So instead of using exp, use the exponential cone construct {x, y, z } == exponential(1) or use rel_entr, as described in mcg’s first answer at Solve optimization problems of exp function .

I don;t guarantee I didn’t miss something in your code.


(ly) #11

Hello ,Stone.Thanks for your advice. I have looked at it.
If x and z are N*M matrices, how should I express it in CVX?


(ly) #12

I added a new constraint where x and z are MN matrices
(Lof./(B
sub_slot))*log(2) + rel_entr(ones(M,N),z) <= 0;

but it is still the same result, I got the result with sdpt3,infeasible with sedumi and mosek


(ly) #13

this is the output of sdpt3
CVX Warning:
Models involving “rel_entr” or other functions in the log, exp, and entropy
family are solved using an experimental successive approximation method.
This method is slower and less reliable than the method CVX employs for
other models. Please see the section of the user’s guide entitled
The successive approximation method
for more details about the approach, and for instructions on how to
suppress this warning message in the future.

Using Pade approximation for exponential
cone with parameters m=3, k=3

Calling SDPT3 4.0: 6603 variables, 2795 equality constraints
For improved efficiency, SDPT3 is solving the dual problem.

num. of constraints = 2795
dim. of sdp var = 3400, num. of sdp blk = 1700
dim. of linear var = 1498
dim. of free var = 5
14 linear variables from unrestricted variable.
*** convert ublk to lblk


SDPT3: Infeasible path-following algorithms


version predcorr gam expon scale_data
HKM 1 0.000 1 0
it pstep dstep pinfeas dinfeas gap prim-obj dual-obj cputime

0|0.000|0.000|1.7e+04|1.0e+00|8.2e+06| 8.793432e+04 0.000000e+00| 0:0:00| spchol 2 2
1|0.000|0.000|1.7e+04|1.0e+00|1.2e+07| 8.498635e+08 -3.140298e-22| 0:0:01| spchol 2 2
2|0.000|0.000|1.7e+04|1.0e+00|2.1e+07| 3.899614e+09 -1.475368e-17| 0:0:01| spchol 2 2
3|0.000|0.000|1.7e+04|1.0e+00|2.9e+07| 8.135212e+09 -3.521631e-16| 0:0:01| spchol 2 2
4|0.000|0.000|1.7e+04|1.0e+00|6.5e+07| 2.875102e+10 -1.215974e-15| 0:0:02|
*** Too many tiny steps: restarting with the following iterate.
*** [X,y,Z] = infeaspt(blk,At,C,b,2,1e5); spchol 2 2
5|0.653|0.527|1.5e+07|1.7e+02|2.6e+13|-5.832740e+12 -4.056442e-13| 0:0:02| spchol 2 2
6|0.140|0.224|1.3e+07|1.3e+02|2.5e+13|-1.709448e+13 -2.322204e-12| 0:0:02| spchol 2 2
7|0.206|0.230|1.0e+07|9.9e+01|2.7e+13|-3.175317e+13 -3.932782e-12| 0:0:03| spchol 2 2
8|0.058|0.095|9.6e+06|8.9e+01|3.0e+13|-5.462450e+13 -1.534045e-11| 0:0:03| spchol 2 2
9|0.277|0.292|7.0e+06|6.3e+01|3.8e+13|-7.020151e+13 -9.458839e-12| 0:0:04| spchol 2 2
10|0.124|0.052|6.1e+06|6.0e+01|6.0e+13|-1.340069e+14 -4.059383e-11| 0:0:04| spchol 2 1
11|0.121|0.353|5.4e+06|3.9e+01|1.2e+14|-1.334580e+14 -1.146392e-10| 0:0:04| spchol 2 2
12|0.180|0.238|4.4e+06|3.0e+01|1.2e+14|-2.462816e+14 -1.968264e-10| 0:0:04| spchol 2 2
13|0.060|0.154|4.1e+06|2.5e+01|1.7e+14|-3.512508e+14 -3.265357e-10| 0:0:05| spchol 2 2
14|0.085|0.204|3.8e+06|2.0e+01|2.3e+14|-4.454187e+14 -6.039578e-10| 0:0:05| spchol 2 2
15|0.090|0.177|3.4e+06|1.6e+01|3.0e+14|-6.493787e+14 -4.510601e-10| 0:0:05| spchol 2 2
16|0.105|0.133|3.1e+06|1.4e+01|4.0e+14|-1.092378e+15 -1.217318e-09| 0:0:06| spchol 2 2
17|0.158|0.195|2.6e+06|1.1e+01|5.9e+14|-1.725419e+15 -2.187015e-09| 0:0:06| spchol 2 2
18|0.188|0.226|2.1e+06|8.9e+00|8.5e+14|-2.212666e+15 -4.294866e-09| 0:0:06| spchol 2 2
19|0.107|0.094|1.9e+06|8.0e+00|1.1e+15|-2.627293e+15 -5.344381e-09| 0:0:07| spchol 2 2
20|0.208|0.419|1.5e+06|4.7e+00|1.3e+15|-3.499767e+15 -1.293871e-08| 0:0:07| spchol 2 2
21|0.065|0.048|1.4e+06|4.4e+00|1.7e+15|-5.860387e+15 -1.549735e-08| 0:0:07| spchol 2 2
22|0.157|0.184|1.2e+06|3.6e+00|3.0e+15|-7.875546e+15 -2.471878e-08| 0:0:07| spchol 2 2
23|0.309|0.174|8.1e+05|3.0e+00|4.4e+15|-1.297810e+16 -4.287900e-08| 0:0:08| spchol 2 2
24|0.218|0.169|6.3e+05|2.5e+00|6.7e+15|-1.169650e+16 -7.561951e-08| 0:0:08| spchol 2 2
25|0.403|0.350|3.8e+05|1.6e+00|6.4e+15|-1.744678e+16 -1.495643e-07| 0:0:09| spchol 2 1
26|0.296|0.356|2.7e+05|1.0e+00|8.2e+15|-1.795220e+16 -3.605018e-07| 0:0:09| spchol 2 2
27|0.308|0.365|1.8e+05|6.6e-01|7.0e+15|-3.068531e+16 -5.991830e-07| 0:0:09| spchol 2 2
28|0.172|0.159|1.5e+05|5.6e-01|1.1e+16|-4.439603e+16 -1.310882e-06| 0:0:10| spchol 2 2
29|0.183|0.337|1.2e+05|3.7e-01|1.7e+16|-6.017116e+16 -1.377684e-06| 0:0:10| spchol 2 2
30|0.124|0.139|1.1e+05|3.2e-01|2.5e+16|-9.807280e+16 -3.621610e-06| 0:0:10| spchol 2 2
31|0.138|0.335|9.4e+04|2.1e-01|4.2e+16|-1.273323e+17 -2.387951e-06| 0:0:10| spchol 2 2
32|0.064|0.042|8.8e+04|2.0e-01|5.7e+16|-1.738033e+17 -5.286437e-06| 0:0:11| spchol 2 2
33|0.139|0.155|7.6e+04|1.7e-01|9.3e+16|-2.620372e+17 -1.644967e-05| 0:0:11| spchol 2 2
34|0.324|0.185|5.1e+04|1.4e-01|1.5e+17|-3.757547e+17 -1.901346e-05| 0:0:11| spchol 2 2
35|0.262|0.442|3.8e+04|7.8e-02|1.8e+17|-3.870359e+17 -4.518464e-05| 0:0:12| spchol 2 2
36|0.171|0.214|3.1e+04|6.1e-02|2.1e+17|-5.516579e+17 -6.508335e-05| 0:0:12| spchol 2 2
37|0.209|0.209|2.5e+04|4.8e-02|3.0e+17|-7.571660e+17 -1.245599e-04| 0:0:12| spchol 2 2
38|0.304|0.348|1.7e+04|3.2e-02|3.6e+17|-9.945712e+17 -2.573597e-04| 0:0:12| spchol 2 2
39|0.282|0.302|1.2e+04|2.2e-02|4.4e+17|-1.396158e+18 -4.885786e-04| 0:0:13| spchol 2 2
40|0.249|0.235|9.3e+03|1.7e-02|6.4e+17|-1.680928e+18 -8.503316e-04| 0:0:13| spchol 2 2
41|0.316|0.373|6.4e+03|1.1e-02|7.5e+17|-2.026566e+18 -1.654018e-03| 0:0:13| spchol 2 2
42|0.265|0.355|4.7e+03|6.8e-03|9.2e+17|-2.364912e+18 -2.976418e-03| 0:0:14| spchol 2 2
43|0.315|0.319|3.2e+03|4.6e-03|1.1e+18|-2.678049e+18 -4.698799e-03| 0:0:14| spchol 2 2
44|0.316|0.343|2.2e+03|3.0e-03|1.3e+18|-2.404346e+18 -7.695738e-03| 0:0:14| spchol 2 2
45|0.461|0.536|1.2e+03|1.4e-03|1.1e+18|-1.636483e+18 -1.301252e-02| 0:0:15| spchol 2 2
46|0.534|0.769|5.5e+02|3.7e-04|7.7e+17|-7.249263e+17 -1.973760e-02| 0:0:15| spchol 2 2
47|0.722|0.554|1.5e+02|1.9e-04|4.1e+17|-5.958667e+16 -2.207967e-02| 0:0:15| spchol 2 2
48|0.713|0.701|4.4e+01|8.2e-05|2.2e+17| 6.757579e+16 -2.564943e-02| 0:0:16| spchol 2 2
49|0.740|0.553|1.1e+01|7.3e-05|9.5e+16| 4.667369e+16 -2.813899e-02| 0:0:16| spchol 2 2
50|0.630|0.789|4.3e+00|4.0e-06|5.3e+16| 3.439021e+16 -3.644068e-02| 0:0:16| spchol 2 2
51|0.722|0.673|1.3e+00|1.3e-06|2.2e+16| 1.637137e+16 -4.286086e-02| 0:0:16| spchol 2 2
52|0.732|0.947|4.3e-01|7.0e-08|9.7e+15| 8.095740e+15 -5.797166e-02| 0:0:17| spchol 2 2
53|0.939|0.979|9.8e-02|3.5e-09|6.6e+14| 5.659861e+14 -5.862012e-02| 0:0:17| spchol 2 2
54|0.979|0.984|7.1e-03|2.8e-07|1.6e+13| 1.220904e+13 -5.862428e-02| 0:0:17| spchol 2 2
55|1.000|0.945|3.6e-04|2.2e-07|8.0e+12| 4.877703e+12 -5.862497e-02| 0:0:18| spchol 1 2
56|0.884|0.976|1.7e-04|6.4e-08|1.2e+12| 7.804334e+11 -5.862498e-02| 0:0:18| spchol 1 1
57|1.000|0.947|8.6e-05|6.4e-08|6.3e+11| 3.896813e+11 -5.862498e-02| 0:0:18| spchol 1 1
58|0.928|0.978|2.9e-05|1.4e-08|6.4e+10| 4.234693e+10 -5.862498e-02| 0:0:19| spchol 1 1
59|0.992|0.968|9.3e-05|9.6e-09|9.2e+09| 6.134139e+09 -5.862498e-02| 0:0:19| spchol 1 1
60|0.983|0.979|4.2e-06|4.4e-09|5.2e+08| 3.910208e+08 -5.862498e-02| 0:0:19| spchol 1 1
61|0.983|0.982|1.2e-06|3.8e-09|2.2e+07| 1.719002e+07 -5.862498e-02| 0:0:19| spchol 1 1
62|0.989|0.985|2.8e-07|3.3e-09|6.0e+05| 4.924154e+05 -5.862498e-02| 0:0:20| spchol 1 1
63|0.989|0.986|3.5e-09|3.2e-09|1.3e+04| 1.071884e+04 -5.862498e-02| 0:0:20| spchol 1 1
64|0.988|0.988|9.8e-11|3.3e-09|1.8e+02| 1.461199e+02 -5.862498e-02| 0:0:20| spchol 1 1
65|0.989|0.989|4.2e-12|3.4e-09|2.0e+00| 1.627087e+00 -5.862498e-02| 0:0:21| spchol 2 2
66|0.989|0.989|4.6e-14|3.2e-09|2.2e-02| 1.789972e-02 -5.862498e-02| 0:0:21| spchol 2 2
67|0.989|0.989|5.0e-16|3.1e-09|2.7e-04| 1.967766e-04 -5.862498e-02| 0:0:21| spchol * 2 * 1
68|0.973|0.989|1.5e-17|3.0e-09|9.8e-06| 7.211932e-06 -5.862498e-02| 0:0:22| spchol * 2 * 1
69|0.983|0.989|7.7e-19|3.1e-09|3.1e-07| 2.047843e-07 -5.862498e-02| 0:0:22| spchol * 1 * 1
70|0.948|0.989|7.1e-19|3.1e-09|1.9e-08| 1.291307e-08 -5.862498e-02| 0:0:22| spchol * 1 * 1
71|0.980|0.989|7.1e-19|3.0e-09|7.2e-10| 4.502536e-10 -5.862498e-02| 0:0:22|
stop: max(relative gap, infeasibilities) < 1.49e-08

number of iterations = 71
primal objective value = 4.50253614e-10
dual objective value = -5.86249785e-02
gap := trace(XZ) = 7.20e-10
relative gap = 6.80e-10
actual relative gap = 5.54e-02
rel. primal infeas (scaled problem) = 7.07e-19
rel. dual " " " = 2.97e-09
rel. primal infeas (unscaled problem) = 0.00e+00
rel. dual " " " = 0.00e+00
norm(X), norm(y), norm(Z) = 1.2e+06, 1.3e+17, 1.3e+17
norm(A), norm(b), norm© = 1.9e+02, 1.0e+00, 2.0e+04
Total CPU time (secs) = 22.41
CPU time per iteration = 0.32
termination code = 0
DIMACS: 7.1e-19 0.0e+00 5.9e-09 0.0e+00 5.5e-02 6.8e-10


Status: Solved
Optimal value (cvx_optval): +0.058625


(Mark L. Stone) #14

You edited your code post from yesterday, such that it no longer has exp. Yet you show output showing the CVXQUAD Pade approximation was invoked. So I don’t even know what you ran. Therefore I can;t diagnose your problem. It may well be that your problem is poorly posed and not reliably solved with any of these methods. Your jumbled and inconsistent mess does not help the prospects for your receiving helpful help.


(Mark L. Stone) #16

I don’t have a definitve answer for you.