A question of sdpvar in Yalmip

[windyy56]

Hi,

Recently, I have got a problem of Yalmip. So I hope people can help me solve the problem. Thank you very much!

This is my code, and the related data are also attached, looking forward your help.

load ins

k=panel(:,1)

e=panel(:,2)

v=panel(:,3)

s=panel(:,4)

r=panel(:,5)

g=panel(:,6)

y=panel(:,7)

b=panel(:,8)

gk=gpanel(:,1)

ge=gpanel(:,2)

gv1=gpanel(:,3)

gs1=gpanel(:,4)

gr=gpanel(:,5)

gg=gpanel(:,6)

gy=gpanel(:,7)

gb=gpanel(:,8)

gv2=gpanel(:,9)

gs2=gpanel(:,10)

H=87

% k e v s r g y b gm1 gm2 ee1 ee2 ee3 ee4

% 1 2 3 4 5 6 7 8 9 10 11 12 13 14

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

z1_result=[]

z2_result=[]

z3_result=[]

z4_result=[]

z5_result=[]

z6_result=[]

z7_result=[]

z8_result=[]

z9_result=[]

z10_result=[]

z11_result=[]

x_result=[]

for i=1:H

x=sdpvar(1,14);

T=x(1)*k(i)+x(2)*e(i)+x(5)*r(i)-x(4)*s(i)+x(9);

M=x(1)*gk(i)+x(2)*ge(i)+x(5)*gr(i)+x(4)*gs2(i);

Q=x(4)*s(i)+x(5)*r(i)+x(3)*v(i)+x(8)*b(i)-x(6)*g(i)-x(7)*y(i)+x(10);

N=x(4)*gs1(i)+x(5)*gr(i)+x(3)*gv1(i)+x(8)*gb(i)+x(6)*gg(i)+x(7)*gy(i);

T1=x(1)*k+x(2)*e+x(5)*r-x(4)*s+x(9)*ones(87,1);

M1=x(1)*gk+x(2)*ge+x(5)*gr+x(4)*gs2;

Q1=x(4)*s+x(5)*r+x(3)*v+x(8)*b-x(6)*g-x(7)*y+x(10)*ones(87,1);

N1=x(4)*gs1+x(5)*gr+x(3)*gv1+x(8)*gb+x(6)*gg+x(7)*gy;

a=T+1;

b=x(11)-M;

c=x(11)+M;

HL=[a;b];

aa=Q+1;

bb=x(12)-N;

cc=x(12)+N;

dd=x(13);

ff=x(14);

dd1=x(9);

ff1=x(10);

HH=[aa;bb];

f=x(11)+x(12)+x(13)+x(14);

F=[norm(HL,2)<=c,norm(HH,2)<=cc,norm(dd1,2)<=dd,norm(ff1,2)<=ff,T1>=zeros(87,1),Q1>=zeros(87,1),x(1)>=0.00000001,x(2)>=0.00000001,x(3)>=0.00000001,x(4)>=0.00000001,x(5)>=0.00000001,x(6)>=0.00000001,x(7)>=0.00000001,x(8)>=0.00000001,T>=0,M>=0,Q>=0,N>=0,T<=1,M<=1,Q<=1,N<=1];

ops=sdpsettings('solver','mosek');

optimize(F,f,ops)

z1=T/M

z2=Q/N

z3=z1+z2

z1_result=[z1_result;value(z1)']

z2_result=[z2_result;value(z2)']

z3_result=[z3_result;value(z3)']

x_result=[x_result;value(x)]

end

By if I run this code in matlab, I always get a mistake as follows. What can I do to solve this problem?

"Error using sdpvar/subsref (line 161)
Index exceeds matrix dimensions.

Error in pins (line 44)
Q=x(4)*s(i)+x(5)*r(i)+x(3)*v(i)+x(8)*b(i)-x(6)*g(i)-x(7)*y(i)+x(10);"

The data file is also provided here.
ins.zip

[johanlofberg]

Does it actually work. Mosek fails on my machine

K>> optimize(F,f,ops)
*** Error(1200): param.param
Return code - 1200 [MSK_RES_ERR_IN_ARGUMENT] [A function argument is incorrect.]

Concerning the error, on second iteration
Q=x(4)*s(i)+x(5)*r(i)+x(3)*v(i)+x(8)*b(i)-x(6)*g(i)-x(7)*y(i)+x(10);
K>> i

i =

 2

K>> b
Linear scalar (real, 5 variables)
Current value: -0.73808
Coefficients range: 0.016762 to 88842.2271

[windyy56]

If I set i is equal to 2\3\4 independently, the code can be solved successfully.

[johanlofberg]

The code runs the first iteration (but has it actually solved the problem or does mosek crash as on my machine?), but then on the second iteration it will fail as you write b(i) where b is a scalar and i=2

[windyy56]

Yes, it is solved in the first iteration in my PC. I have used the related code in many cases. So I feel confused how to change the code to run it successfully? Could you please give me the corrent expresion in Q=x(4)*s(i)+x(5)*r(i)+x(3)*v(i)+x(8)*b(i)-x(6)*g(i)-x(7)*y(i)+x(10); or in other places. In my opinon, b(2) is a number and it can be used.

[johanlofberg]

You define b as a scalar inside the loop

[windyy56]

Got it, I have use "b" twice in this code but in different meanings. I also define it as b=x(11)-M. Thus the conflict appears. Thanks very much for your help!