Following my previous topic on how to optimize a complex function, I decided to write a custom algorithm because I need this task to be run locally and on machines with no access to Mathematica.
Unfortunately the performance differences between my java code and the Mathematica's one are really huge and I need another solution.
I've just discovered Wolfram Cloud and this seems the best way to make everything work because I can just execute my Mathematica's notebook, so I subscribed to the free tier to give it a try.
Here's what I'm trying:
optimize[d_, p_] := Module[{n, m, q, k, A, b, result},
n=Length[d];
m=Dimensions[p][[2]];
vec[m_]:=Flatten[m\[Transpose]];
mat[v_]:=Partition[v,n]\[Transpose];
j[k_]:=ConstantArray[1,k];
q=j[m];
k=vec[p (d\[TensorProduct]q)];
A=ArrayFlatten[{IdentityMatrix[m]\[TensorProduct]j[n]}];
b=j[m]\[TensorProduct]{1,0};
result = mat[LinearProgramming[k,A,b]];
result
];
CloudDeploy[APIFunction[{"d" -> "List", "p" -> "List"}, optimize[#d, #p] &, "String"]]
I'm using List as a parameter because d and p are like
d={7,5,11,3};
p=({{5,2,5,3,7,1,1,5,2,5,3,7,1,1},{2,8,3,2,7,2,3,2,8,3,2,7,2,3},{3,6,4,5,9,1,1,3,6,4,5,9,1,1},{4,6,2,8,14,8,4,4,6,2,8,14,8,4}});
Now, the function works fine in my local notebook and hitting evaluate actually makes the deploy.
The problem is that the API doesn't return what I expect.
While I expect a string that contains the result matrix, it outputs:
Transpose[LinearProgramming[LinearProgramming[Transpose[{{{5,2,5,3,7,1,1,5,2,5,3,7,1,1},{2,8,3,2,7,2,3,2,8,3,2,7,2,3},{3,6,4,5,9,1,1,3,6,4,5,9,1,1},{4,6,2,8,14,8,4,4,6,2,8,14,8,4}} {{7,5,11,3}} ? ConstantArray[1, {1}[[2]]]}]], LinearProgramming[ArrayFlatten[{IdentityMatrix[{1}[[2]]] ? {1}}]], LinearProgramming[ConstantArray[1, {1}[[2]]] ? {1, 0}]]]
and I don't understand why.
Do you know what the problem is? Their documentation is not that helpful and I'm quite clueless.
