error in running my optimization code

My code is throwing the following error in conversion which i cannot understand.

Code-

yexpected=table2array(Book65);
tspan=0:48;
yo=[0.2 0.05 0.539];
p=optimvar('p',7,"LowerBound",0,"UpperBound",30);
sol=ode45(@(t,y)kumar2004optimizer(t,y,p),tspan,yo);
myfcn = fcn2optimexpr(@kumar2004optimizer,p,tspan,y0);
obj = sum(sum((myfcn - yexpected).^2));
prob = optimproblem("Objective",obj);
p0.p = [3 25 25 12 1 0.5 1];
[psol,sumsq] = solve(prob,p0);
function dydt=kumar2004optimizer(t,y,p)
dydt=zeros(3,1);
dydt(1)=p(1)*y(1)*(1-y(1))-p(2)*y(1)*y(2);
dydt(2)=(p(3)*y(1)+y(3))*y(2)*(1-y(2))-y(2);
dydyt(3)=p(4)*(1+tanh((y(2)-p(5))/p(6)))-p(7)*y(3);
end

error-Unable to perform assignment because value of type

'optim.problemdef.OptimizationExpression' is not convertible to 'double'.

Answer 1

[Torsten]

I assume that "yexpected" is a matrix of size 49x3.

yexpected=table2array(Book65);
p0 = [3 25 25 12 1 0.5 1];
sol = lsqnonlin(@(p)optimizer(p,yexpected),p0)

function res = optimizer(p,yexpected)
  tspan=0:48;
  y0=[0.2 0.05 0.539];
  sol=ode45(@(t,y)kumar2004optimizer(t,y,p),tspan,y0);
  res = sol - yexpected;
  res = res(:);
end

function dydt=kumar2004optimizer(t,y,p)
  dydt=zeros(3,1);
  dydt(1)=p(1)*y(1)*(1-y(1))-p(2)*y(1)*y(2);
  dydt(2)=(p(3)*y(1)+y(3))*y(2)*(1-y(2))-y(2);
  dydt(3)=p(4)*(1+tanh((y(2)-p(5))/p(6)))-p(7)*y(3);
end

[Ayush Ranjan]

I tried running this but it says I cannot use '-' operator for structure

[Torsten]

"yexpected" must be a numerical array of size 49x3. See how you get "Book65" transformed to this class and size.

[Ayush Ranjan]

I changed it to array but it shows the following error.

Operator '-' is not supported for operands of type 'struct'.

Error in optimizertrial>optimizer (line 8)

res = sol - yexpected;

Error in optimizertrial>@(p)optimizer(p,yexpected) (line 3)

sol = lsqnonlin(@(p)optimizer(p,yexpected),p0);

Error in lsqnonlin (line 218)

initVals.F = feval(funfcn{3},xCurrent,varargin{:});

Error in optimizertrial (line 3)

sol = lsqnonlin(@(p)optimizer(p,yexpected),p0);

[Torsten]

Before you pass "yexpected" to "lsqnonlin" by the line

sol = lsqnonlin(@(p)optimizer(p,yexpected),p0)

add the commands

class(yexpected)
size(yexpected)

and show us the output.

[Ayush Ranjan]

ans =

'double'

ans =

26 3

This is what I got on the above commands

[Torsten]

What are the times where these 26x3 measurements were taken ? You have to specifiy them in "tspan" in place of the 0:48 you use at the moment.

And (I made a mistake here) use

[T,Y]=ode45(@(t,y)kumar2004optimizer(t,y,p),tspan,y0);
res = Y - yexpected;

instead of

sol=ode45(@(t,y)kumar2004optimizer(t,y,p),tspan,y0);
res = sol - yexpected;

[Ayush Ranjan]

it worked thank you