TEST(Services, do_bfgs_optimize__lbfgs) {
std::vector<double> cont_vector(2);
cont_vector[0] = -1; cont_vector[1] = 1;
std::vector<int> disc_vector;
static const std::string DATA("");
std::stringstream data_stream(DATA);
stan::io::dump dummy_context(data_stream);
Model model(dummy_context);
typedef stan::optimization::BFGSLineSearch<Model,stan::optimization::LBFGSUpdate<> > Optimizer_LBFGS;
Optimizer_LBFGS lbfgs(model, cont_vector, disc_vector, &std::cout);
double lp = 0;
bool save_iterations = true;
int refresh = 0;
int return_code;
unsigned int random_seed = 0;
rng_t base_rng(random_seed);
std::fstream* output_stream = 0;
mock_callback callback;
return_code = stan::services::optimization::do_bfgs_optimize(model, lbfgs, base_rng,
lp, cont_vector, disc_vector,
output_stream, &std::cout,
save_iterations, refresh,
callback);
EXPECT_FLOAT_EQ(return_code, 0);
EXPECT_EQ(35, callback.n);
}
TEST(StanCommon, reject_func_call_generated_quantities) {
std::string error_msg = "user-specified rejection";
std::fstream empty_data_stream(std::string("").c_str());
stan::io::dump empty_data_context(empty_data_stream);
empty_data_stream.close();
std::stringstream model_output;
model_output.str("");
// instantiate model
reject_func_call_generated_quantities_model_namespace::reject_func_call_generated_quantities_model* model
= new reject_func_call_generated_quantities_model_namespace::reject_func_call_generated_quantities_model(empty_data_context, &model_output);
// instantiate args to log_prob function
Eigen::VectorXd cont_params = Eigen::VectorXd::Zero(model->num_params_r());
std::vector<double> cont_vector(cont_params.size());
for (int i = 0; i < cont_params.size(); ++i)
cont_vector.at(i) = cont_params(i);
std::vector<int> disc_vector;
double lp;
boost::ecuyer1988 base_rng;
base_rng.seed(123456);
lp = model->log_prob<false, false>(cont_vector, disc_vector, &std::cout);
try {
stan::services::io::write_iteration(model_output, *model, base_rng,
lp, cont_vector, disc_vector);
} catch (const std::domain_error& e) {
if (std::string(e.what()).find(error_msg) == std::string::npos) {
FAIL() << std::endl << "---------------------------------" << std::endl
<< "--- EXPECTED: error_msg=" << error_msg << std::endl
<< "--- FOUND: e.what()=" << e.what() << std::endl
<< "--------------------------------*" << std::endl
<< std::endl;
}
return;
}
FAIL() << "model failed to do reject" << std::endl;
}
/**
* Runs the algorithm and writes to output.
*
* @param tol_rel_obj relative tolerance parameter for convergence
* @param max_iterations max number of iterations to run algorithm
*/
int run(double tol_rel_obj, int max_iterations) const {
if (diag_stream_) {
*diag_stream_ << "iter,time_in_seconds,ELBO" << std::endl;
}
// initialize variational approximation
Q variational = Q(cont_params_);
// run inference algorithm
robbins_monro_adagrad(variational, tol_rel_obj, max_iterations);
// get mean of posterior approximation and write on first output line
cont_params_ = variational.mean();
std::vector<double> cont_vector(cont_params_.size());
for (int i = 0; i < cont_params_.size(); ++i)
cont_vector.at(i) = cont_params_(i);
std::vector<int> disc_vector;
if (out_stream_) {
services::io::write_iteration(*out_stream_, model_, rng_,
0.0, cont_vector, disc_vector,
print_stream_);
}
// draw more samples from posterior and write on subsequent lines
if (out_stream_) {
for (int n = 0; n < n_posterior_samples_; ++n) {
cont_params_ = variational.sample(rng_);
for (int i = 0; i < cont_params_.size(); ++i)
cont_vector.at(i) = cont_params_(i);
services::io::write_iteration(*out_stream_, model_, rng_,
0.0, cont_vector, disc_vector, print_stream_);
}
}
return stan::services::error_codes::OK;
}
