"DOUBLE cannot convert the input expression into a double array. If the input expression contains a symbolic variable, use the VPA function instead."

My command window code is:

 syms L C gamma lambda k K X U FOG SOG u1 u2 u3 eps1 eps2;
C=u1^2+3*u2^2+2*u3^2;
L = u1*u2+2*u1*u3+3*u2*u3;
U=[u1 u2];X=[u3];
C=u1^2+3*u2^2+2*u3^2;
gamma=4;
Cx=jacobian(C,X);
Cu=jacobian(C,U);
Lu=jacobian(L,U);
Lx=jacobian(L,X);
lambda_T=-Lx*inv(Cx);
Lbar_u=Lu+lambda_T*Cu;
Lbar_x=Lx+lambda_T*Cx;
Lbar_ux=jacobian(Lbar_u,X);
Lbar_xx=jacobian(Lbar_x,X);
Lbar_xu=jacobian(Lbar_x,U);
Lbar_uu=jacobian(Lbar_u,U);
Lxu=jacobian(Lx,U);
Lux=jacobian(Lu,X);
Lbar_uu_star=Lbar_uu-Lbar_ux*inv(Cx)*Cu-(Cu)'*((Cx)')^(-1)*Lbar_xu+(Cu)'*((Cx)')^(-1)*Lbar_xx*inv(Cx)*Cu;
u3=-1;u1=1;u2=1;
k=1;K=.500330;
FOG=0;SOG=1;
Opt(0,K,1,Lbar_uu_star,L,X,U,C,k,gamma,lambda_T,Cx,Cu,Lx,Lu,Lxu,Lux,Lbar_u,Lbar_uu,Lbar_x,Lbar_xu, Lbar_xx, Lbar_ux,eps1,eps2);

My M-file Opt.m is:

function[] = Opt(FOG,K,SOG,Lbar_uu_star,L,X,U,C,k,gamma,lambda_T,Cx,Cu,Lx,Lu,Lxu,Lux,Lbar_u,Lbar_uu,Lbar_x,Lbar_xu, Lbar_xx, Lbar_ux,eps1,eps2)

while abs(double(subs(C))-double(subs(gamma)))>=(double(subs(eps1)))

X=(double(subs(X)))-inv(double(subs(Cx)))*(double(subs(C))-(double(subs(gamma))));

if abs(double(subs(C))-(double(subs(gamma))))<(double(subs(eps1)))

if abs(double(subs(Lbar_u)))<(double(subs(eps2)))

break;

else

k=k+1;

if FOG == 1

U=(double(subs(U)))-K*(double(subs((Lbar_u)')));

elseif SOG == 1

U=(double(subs(U)))-inv(double(subs(Lbar_uu_star)))*(double(subs(Lbar_u)));

end

end

end

end

L_star=((double(subs(X^2/2))))+(double(subs((U^2)/2)))-2*(double(subs(X)))-2*(double(subs(U)));

X_star=double(subs(X));

U_star=double(subs(U));

if FOG == 1

    display('First order gradient method')

elseif SOG == 1

    display('Second order gradient method')

end

display(X_star);display(U_star);display(L_star);display(k);

end

errors I get after running the code are:

Error using mupadmex Error in MuPAD command: DOUBLE cannot convert the input expression into a double array.

If the input expression contains a symbolic variable, use the VPA function instead.

Error in sym/double (line 702) Xstr = mupadmex('symobj::double', S.s, 0);

Error in Opt (line 3) while abs(double(subs(C))-double(subs(gamma)))>=(double(subs(eps1)))

I have no Idea what is going on. Any help would be highly appreciated

Answer 1

[Carlos]

The problem comes in

double(subs(C))

From your code

C=u1^2+3*u2^2+2*u3^2;

so when you do

subs(C)

ans =

u1^2 + 3*u2^2 + 2*u3^2

C remains unchanged, so you cannot convert a symbol to a number.

If you did

    subs(C,{u1,u2,u3},{1,1,1})

ans =

       6

So in conclusion you cannot convert a symbolic expression containing variables into a number.

[Walter Roberson]

Correct. I have said the same thing in past postings to upinderjeet: it is not possible to loop with a condition testing symbolic variables that have no numeric values.

[Carlos]

Something I don´t uderstand about the code is what upinderjeet tries to achieve by doing

subs(C)

Exactly Walter it is not possible,in fact you taught me in past questions how to use symbolic variables.

[upinderjeet]

Walter and Carlos

Thanks first of all for the quick reply. Actually If you notice the while loop the value of u1,u2 and u3 is getting changed if the certain conditions are met. So I want "C" to acquire the latest value of itself according to the updated values of u1,u2 and u3. So by using "double(subs())" command I want to accomplish that and actually everything worked fine at the end for my previous issues on which walter was helping me but this is similar and related problem but the only differnce is that X and U now are vector parameters unlike the scalar case of the past Walter.

[Walter Roberson]

When you subs() and expect symbolic variables to be replaced by what is in the workspace, the substitution is limited to variables existing in the current workspace. In your code, u1 and so on are defined in the caller of the routine that does the subs()