How do I convert an armadillo matrix to a vector of vectors?

In such cases you should use arma::conv_to which is a totally superb feature of arma.

Note that this method will require from a source object to be able to be interpreted as a vector. That is why we need to do this iteratively for every row. Here is a conversion method:

#include <armadillo>

typedef std::vector<double> stdvec;
typedef std::vector< std::vector<double> > stdvecvec;

stdvecvec mat_to_std_vec(arma::mat &A) {
    stdvecvec V(A.n_rows);
    for (size_t i = 0; i < A.n_rows; ++i) {
        V[i] = arma::conv_to< stdvec >::from(A.row(i));
    };
    return V;
}

And here is an exemplary usage:

#include <iomanip>
#include <iostream>

int main(int argc, char **argv) {
    arma::mat A = arma::randu<arma::mat>(5, 5);
    std::cout << A << std::endl;

    stdvecvec V = mat_to_std_vec(A);
    for (size_t i = 0; i < V.size(); ++i) {
        for (size_t j = 0; j < V[i].size(); ++j) {
            std::cout << "   "
                << std::fixed << std::setprecision(4) << V[i][j];
        }
        std::cout << std::endl;
    }
    return 0;
}

std::setprecision used to generate more readable output:

0.8402   0.1976   0.4774   0.9162   0.0163
0.3944   0.3352   0.6289   0.6357   0.2429
0.7831   0.7682   0.3648   0.7173   0.1372
0.7984   0.2778   0.5134   0.1416   0.8042
0.9116   0.5540   0.9522   0.6070   0.1567

0.8402   0.1976   0.4774   0.9162   0.0163
0.3944   0.3352   0.6289   0.6357   0.2429
0.7831   0.7682   0.3648   0.7173   0.1372
0.7984   0.2778   0.5134   0.1416   0.8042
0.9116   0.5540   0.9522   0.6070   0.1567

Have a good one!