What is the best practice about using CppAD in Monte Carlo?

[xuruilong100]

To get Greeks of exotic options, I'm trying to use CppAD in Monte Carlo simulation. Below is my idea of computing pathwise Greeks.

...

double s0 = 100.0;
double vol = 0.1;

std::vector<AD<double>> adParameters(2);
adParameters[0] = s0;
adParameters[1] = vol;

std::vector<AD<double>> adOpt(1);

std::vector<double> jac(2);
std::vector<double> greeks(2, 0.0);
std::vector<double> parameters(2);
parameters[0] = s0;
parameters[1] = vol;

for (size_t i = 0; i < samples; ++i){
    CppAD::Independent(adParameters);

    // step 1: generate a path by parameters
    auto path = PathMaker(adParameters);

    // step 2: get option price of one path 
    adOpt[0] = PathPricer(path);

    CppAD::ADFun<double> f(adParameters, adOpt);
    
    jac = f.Jacobian(parameters);

    greeks[0] += jac[0];
    greeks[1] += jac[1];
}

greeks[0] /= samples;
greeks[1] /= samples;

return greeks;

IS IT RIGHT? And, what shall I do to improve performance?

[bradbell]

I suggest you take a look at the following discussion:
#235 (reply in thread)