rmontanana/BayesNet
1
0
Fork 0
Code Issues 2 Pull Requests 1 Actions Packages Projects Releases 6 Wiki Activity

Fix Experiment

Browse Source
This commit is contained in:
Ricardo Montañana Gómez
2023-07-26 14:11:49 +02:00
parent af7a1d2b40
commit 49a49a9dcd
2 changed files with 26 additions and 16 deletions
Download Patch File Download Diff File Expand all files Collapse all files

31
src/Platform/Experiment.cc
View File

@@ -25,9 +25,11 @@ Result cross_validation(Fold* fold, string model_name, Tensor& X, Tensor& y, vec
{ "SPODE", new bayesnet::SPODE(2) }, { "TAN", new bayesnet::TAN() }
}
);
auto Xt = torch::transpose(X, 0, 1);
auto result = Result();
auto k = fold->getNumberOfFolds();
auto accuracy = torch::zeros({ k }, kFloat64);
auto accuracy_test = torch::zeros({ k }, kFloat64);
auto accuracy_train = torch::zeros({ k }, kFloat64);
auto train_time = torch::zeros({ k }, kFloat64);
auto test_time = torch::zeros({ k }, kFloat64);
Timer train_timer, test_timer;
@@ -37,9 +39,9 @@ Result cross_validation(Fold* fold, string model_name, Tensor& X, Tensor& y, vec
auto [train, test] = fold->getFold(i);
auto train_t = torch::tensor(train);
auto test_t = torch::tensor(test);
auto X_train = X.index({ train_t, "..." });
auto X_train = Xt.index({ "...", train_t });
auto y_train = y.index({ train_t });
auto X_test = X.index({ test_t, "..." });
auto X_test = Xt.index({ "...", test_t });
auto y_test = y.index({ test_t });
model->fit(X_train, y_train, features, className, states);
cout << "Training Fold " << i + 1 << endl;
@@ -48,12 +50,15 @@ Result cross_validation(Fold* fold, string model_name, Tensor& X, Tensor& y, vec
cout << "X_test: " << X_test.sizes() << endl;
cout << "y_test: " << y_test.sizes() << endl;
train_time[i] = train_timer.getDuration();
auto accuracy_train_value = model->score(X_train, y_train);
test_timer.start();
auto acc = model->score(X_test, y_test);
auto accuracy_test_value = model->score(X_test, y_test);
test_time[i] = test_timer.getDuration();
accuracy[i] = acc;
accuracy_train[i] = accuracy_train_value;
accuracy_test[i] = accuracy_test_value;
}
result.setScore(torch::mean(accuracy).item<double>());
result.setScoreTest(torch::mean(accuracy_test).item<double>()).setScoreTrain(torch::mean(accuracy_train).item<double>());
result.setScoreTestStd(torch::std(accuracy_test).item<double>()).setScoreTrainStd(torch::std(accuracy_train).item<double>());
result.setTrainTime(torch::mean(train_time).item<double>()).setTestTime(torch::mean(test_time).item<double>());
return result;
}
@@ -114,8 +119,9 @@ int main(int argc, char** argv)
catch (...) {
throw runtime_error("Number of folds must be an integer");
}});
program.add_argument("-s", "--seed").help("Random seed").default_value(-1).scan<'i', int>();
bool class_last, discretize_dataset, stratified;
int n_folds;
int n_folds, seed;
string model_name, file_name, path, complete_file_name;
try {
program.parse_args(argc, argv);
@@ -125,6 +131,7 @@ int main(int argc, char** argv)
discretize_dataset = program.get<bool>("discretize");
stratified = program.get<bool>("stratified");
n_folds = program.get<int>("folds");
seed = program.get<int>("seed");
complete_file_name = path + file_name + ".arff";
class_last = datasets[file_name];
if (!file_exists(complete_file_name)) {
@@ -142,18 +149,16 @@ int main(int argc, char** argv)
auto [X, y, features, className, states] = loadDataset(path, file_name, class_last, discretize_dataset);
Fold* fold;
if (stratified)
fold = new StratifiedKFold(n_folds, y, -1);
fold = new StratifiedKFold(n_folds, y, seed);
else
fold = new KFold(n_folds, y.numel(), -1);
fold = new KFold(n_folds, y.numel(), seed);
auto experiment = Experiment();
experiment.setDiscretized(discretize_dataset).setModel(model_name).setPlatform("cpp");
experiment.setStratified(stratified).setNFolds(5).addRandomSeed(271).setScoreName("accuracy");
experiment.setStratified(stratified).setNFolds(n_folds).addRandomSeed(seed).setScoreName("accuracy");
auto result = cross_validation(fold, model_name, X, y, features, className, states);
result.setDataset(file_name);
experiment.addResult(result);
experiment.save(path);
for (auto& item : states) {
cout << item.first << ": " << item.second.size() << endl;
}
experiment.show();
return 0;
}