Function main

examples/least_squares_ex.cpp:89–180

// ----------------------------------------------------------------------------------------

int main()
{
    try
    {
        // randomly pick a set of parameters to use in this example
        const parameter_vector params = 10*randm(3,1);
        cout << "params: " << trans(params) << endl;


        // Now let's generate a bunch of input/output pairs according to our model.
        std::vector<std::pair<input_vector, double> > data_samples;
        input_vector input;
        for (int i = 0; i < 1000; ++i)
        {
            input = 10*randm(2,1);
            const double output = model(input, params);

            // save the pair
            data_samples.push_back(make_pair(input, output));
        }

        // Before we do anything, let's make sure that our derivative function defined above matches
        // the approximate derivative computed using central differences (via derivative()).  
        // If this value is big then it means we probably typed the derivative function incorrectly.
        cout << "derivative error: " << length(residual_derivative(data_samples[0], params) - 
                                               derivative(residual)(data_samples[0], params) ) << endl;





        // Now let's use the solve_least_squares_lm() routine to figure out what the
        // parameters are based on just the data_samples.
        parameter_vector x;
        x = 1;

        cout << "Use Levenberg-Marquardt" << endl;
        // Use the Levenberg-Marquardt method to determine the parameters which
        // minimize the sum of all squared residuals.
        solve_least_squares_lm(objective_delta_stop_strategy(1e-7).be_verbose(), 
                               residual,
                               residual_derivative,
                               data_samples,
                               x);

        // Now x contains the solution.  If everything worked it will be equal to params.
        cout << "inferred parameters: "<< trans(x) << endl;
        cout << "solution error:      "<< length(x - params) << endl;
        cout << endl;




        x = 1;
        cout << "Use Levenberg-Marquardt, approximate derivatives" << endl;
        // If we didn't create the residual_derivative function then we could
        // have used this method which numerically approximates the derivatives for you.
        solve_least_squares_lm(objective_delta_stop_strategy(1e-7).be_verbose(),