int parameter_server(int argc, char *argv[])
{
std::vector<std::vector<double> > training; // TODO: can be optimized, we can count the number of feats and allocate accordingly
std::vector<std::vector<double> > validation; // TODO: can be optimized, we can count the number of feats and allocate accordingly
std::unordered_set<double> label_set;
std::unordered_map<double, std::vector<double> > model;
char *training_file = argv[1];
assert(strlen(argv[2]) == 1);
char delimiter = argv[2][0];
// read training data
std::cout << "Reading training data ... ";
read_training_data(training_file, delimiter, training, validation, label_set);
std::cout << "[Done]" << std::endl;
int data_passes = std::stoi(std::string(argv[6]));;
size_t batch_size = training.size();
if (argc == 8)
batch_size = std::stoul(std::string(argv[7]));
std::cout << std::endl
<< "***Info***" << std::endl
<< "Training Data File: " << training_file << std::endl
<< "Learning Rate: " << learning_rate << std::endl
<< "Regularization Parameter: " << reg_param << std::endl
<< "Data Passes: " << data_passes << std::endl
<< "Batch Size: " << batch_size << std::endl
<< "Num Labels: " << label_set.size() << std::endl
<< "Training Set Size: " << training.size() << std::endl
<< "Validation Set Size: " << validation.size() << std::endl
<< "**********" << std::endl;
// logistic regression
logistic_regression(training, label_set, model, learning_rate, reg_param, batch_size, data_passes);
// print the f-score(s)
const std::unordered_map<double, double> fsc = fscore(validation, label_set, model);
std::cout << std::endl << std::endl;
std::cout << "F-Score(s):" << std::endl;
std::cout << "Label\tScore" << std::endl;
for (std::unordered_map<double, double>::const_iterator it = fsc.begin(); it != fsc.end(); ++it) {
std::cout << it->first << "\t" << it->second << std::endl;
}
return 0;
}
//--------------------------------------------------------------------
int main(int argc, char* argv[])
{
t_parameters params;
params.num_sub_populations = 2;
params.sub_population_size = 30; // the number of individuals in population (must be an even number!)
params.code_length = 50;
params.num_generations = 1000; // the number of generations
params.mutation_probability = 0.01; // mutation probability
params.crossover_probability = 0.9; // crossover probability
params.variables_probability = 0.4;
params.operators_probability = 0.5;
params.constants_probability = 1 - params.variables_probability - params.operators_probability; // sum of variables_prob + operators_prob + constants_prob MUST BE 1 !
params.num_constants = 10; // use 3 constants from -1 ... +1 interval
params.constants_min = -1;
params.constants_max = 1;
#ifdef USE_THREADS
params.num_threads = 4;
#endif
t_graph *graphs = NULL;
int num_graphs = 0;
if (!read_training_data(graphs, num_graphs)) {
printf("Cannot find input file(s)! Please specify the full path!\n");
printf("Press Enter ...");
getchar();
return 1;
}
compute_global_variables(graphs, num_graphs);
int num_variables = 10;
int current_proc_id = 0;
int num_procs = 0;
#ifdef USE_MPI
MPI_Init(&argc, &argv);
MPI_Comm_size(MPI_COMM_WORLD, &num_procs);
MPI_Comm_rank(MPI_COMM_WORLD, ¤t_proc_id);
#endif
srand(current_proc_id); // we run each process with a different seed
printf("Evolving... proc_id = %d\n", current_proc_id);
start_steady_state(params, graphs, 2, graphs + 2, 2, num_variables, num_procs, current_proc_id);
delete_training_graphs(graphs, num_graphs);
#ifdef USE_MPI
if (current_proc_id == 0) {
printf("Press Enter ...");
getchar();
}
MPI_Finalize();
#endif
return 0;
}
