oxfordcontrol / cdcs Goto Github PK

Hi,

Thanks for your work.

I am trying to check the feasibility of some LMIs. During the iteration progress of CDCS, all the elements show as NaN. One instance is attached (Matlab code, but save as txt to upload).
Code_for_Seuret2013AutomaticaCase3.txt

I also tried to check the feasibilities of some simple Lyapunov inequalities, and they were successfully solved.

My question is: can the instance be solved by CDCS? If so, how to fix my code.

Thanks

I get the following error message if K.q = []:

Operands to the || and && operators must be convertible to logical scalar
values.

Error in cdcs_utils.checkInputs (line 40)
if (isfield(K,'q') && max(K.q) > 0)

Error in cdcs (line 104)
[At,b,c,K,opts] = checkInputs(At,b,c,K,opts);

Of course I can get around that issue by deleting the field q from K before calling cdcs.
However, it may be more user friendly to check that K.q is nonempty in checkInputs:

isfield(K,'q') && ~isempty(K.q) && max(K.q) > 0

Hi, I have a question regarding the function flatten in CDCS/packages/+cdcs_utils/flatten.m.

I think I understand what it is supposed to do; namely take a cell array containing arrays of various sizes, flatten each one of them and stack them up in a big column.

The part that I am confused about is the inputs of this function.

The function takes in 3 different aguments, lower case z, upper case Z and a flag svecFlag that is used to call CDCS/packages/+cdcs_utils/private/svec.c.
However, I don't see the first argument (lower case z) being used anywhere.

The last line of the function creates a lower case z that is the returned object but as far as I understand this is not per say using the z passed as input.

Is there something I am missing ?
Thanks a lot for any feedback :)

How does CDCS interpret a nonsymmetric SDP matrix ? For instance, if in the dual form,
A'y gives the matrix

[1 x
 y 1]

what does it do ? Does it interpret it as

[1 (x+y)/2
 (x+y)/2 1]

should be PSD (like SeDuMi does)
or that

[1 x
 x 1]

should be PSD and y should be equal to x

CDCS stops after maxIter iterations and returns a vector X that contains NaNs when I try to solve the following problem instance: prob.zip

I observe the same problem with many other problem instances. Could this be a problem with my installation or a bug in CDCS? I'm using the most recent version of CDCS from this repo with MATLAB R2017a on macOS.

clear all
load prob.mat
[x,y,sedumi_info] = sedumi(A,b,c,K);

opts = cdcsOpts;
opts.maxIter = 5000;
[xt,yt,zt,cdcs_info] = cdcs(A,b,c,K,opts);

any(isnan(xt))