Clustering Iris

// Iris: Fisher, R. (1936). UCI. https://doi.org/10.24432/C56C76
// CC BY 4.0: https://creativecommons.org/licenses/by/4.0/
// Original UCI records retained; class names mapped to zero-based IDs.
// Full attribution and variant notes: datasets/iris/README.md.
import aner.dataset;
import aner.tensor;
import aner.ml;
import aner.metrics;

fn main() -> Unit {
    let flowers = dataset.iris();
    let split_seed = 2026;
    let partitions = dataset.split(flowers, 0.2, split_seed);
    let training = dataset.train(partitions);
    let testing = dataset.test(partitions);
    // Fit preprocessing only on training rows; reuse it unchanged for test rows.
    let scaler = dataset.fit_standardizer(dataset.features(training));
    let x_train = dataset.transform(scaler, dataset.features(training));
    let x_test = dataset.transform(scaler, dataset.features(testing));
    let test_labels = dataset.targets(testing);

    print("Dataset citation"); print(dataset.citation(flowers));
    print("Dataset license"); print(dataset.license(flowers));
    print("Dataset version"); print(dataset.version(flowers));
    print("Dataset SHA256"); print(dataset.sha256(flowers));
    print("Split seed"); print(split_seed);
    print("Training rows"); print(dataset.rows(training));
    print("Test rows"); print(dataset.rows(testing));

    // Fit on features only. Class labels are used below only for evaluation.
    // One fixed initialization is educational; it does not find a guaranteed optimum.
    let clusters = 3;
    let initialization_seed = 42;
    let model: KMeansModel = ml.fit_kmeans(x_train, clusters, 100, 0.000001, initialization_seed);
    print("Clusters"); print(clusters);
    print("Initialization seed"); print(initialization_seed);
    print("Converged"); print(ml.converged(model));
    print("Iterations"); print(ml.iterations(model));
    print("Training inertia (sum of squared distances)"); print(ml.inertia(model));
    let assignments = ml.predict(model, x_test);
    // Cluster numbers have no class meaning: ARI compares partitions up to renaming.
    print("Test adjusted Rand index");
    print(metrics.adjusted_rand_index(assignments, test_labels));
    print("Centers in standardized feature coordinates; row-major values");
    let centers = ml.centers(model);
    var row = 0;
    while row < tensor.rows(centers) {
        var col = 0;
        while col < tensor.cols(centers) {
            print(tensor.value(centers, row, col));
            col = col + 1;
        }
        row = row + 1;
    }
}