A optimizer framework written in my spare time.

opt::Variable x = opt::AddVariable("x");
opt::Constant e = 2.71828_const;
opt::Expression expr = (x^4)*(e^-x);

To evaluate expression

// it will use the current value of variables
cout<<expr.eval()<<endl;

Pack expression into function(aka define function)

// build f(x)
auto f = opt::MakeFunction(expr, x);
cout<<f(1.0)<<endl;

Constant literal

opt::Constant pi = 3.1415926_const;

You must install numpy and matplotlib first to draw graph. For some unknown reason, I cannot get numpy work in python_d. Thus, debug build using VC++ may not run correctly. However, mingw and release mode works perfectly. Note that if you are using mingw, you might need to modify pyconfig.h

pip install numpy matplotlib

mkdir build && cd build && cmake .. -G "Visual Studio 15 2017 Win64" -DENABLE_PLOT=1
cmake --build . --config Release

Remove this line in pyconfig.h

#define hypot _hypot

Because of the natural of the two search algorithm, Golden and fib search failed to find correct minima within given [a, b] interval.

>p1.exe
9.99999
10

I combined univaridate algorithm(cyclic coordinate) and Powell's method with golden search. Since there are two minima in the objective function, I called both function twice with different initial points.

>p2.exe
Univaridate method:
-1.00043 -0.501071
1.00043 0.501068
Powell's method:
-1 -0.500003
1 0.500003