sklearn-docs/optics_clustering

1# Scikit learn example https://scikit-learn.org/stable/auto_examples/cluster/plot_optics.html2 3import gradio as gr4 5from sklearn.cluster import OPTICS, cluster_optics_dbscan6import matplotlib.gridspec as gridspec7import matplotlib.pyplot as plt8import numpy as np9 10plt.switch_backend("agg")11 12# Theme from - https://huggingface.co/spaces/trl-lib/stack-llama/blob/main/app.py13theme = gr.themes.Monochrome(14    primary_hue="indigo",15    secondary_hue="blue",16    neutral_hue="slate",17    radius_size=gr.themes.sizes.radius_sm,18    font=[19        gr.themes.GoogleFont("Open Sans"),20        "ui-sans-serif",21        "system-ui",22        "sans-serif",23    ],24)25 26 27def do_submit(n_points_per_cluster, min_samples, xi, min_cluster_size):28    # # Generate sample data29    np.random.seed(0)30    n_points_per_cluster = int(n_points_per_cluster)31 32    C1 = [-5, -2] + 0.8 * np.random.randn(n_points_per_cluster, 2)33    C2 = [4, -1] + 0.1 * np.random.randn(n_points_per_cluster, 2)34    C3 = [1, -2] + 0.2 * np.random.randn(n_points_per_cluster, 2)35    C4 = [-2, 3] + 0.3 * np.random.randn(n_points_per_cluster, 2)36    C5 = [3, -2] + 1.6 * np.random.randn(n_points_per_cluster, 2)37    C6 = [5, 6] + 2 * np.random.randn(n_points_per_cluster, 2)38    X = np.vstack((C1, C2, C3, C4, C5, C6))39 40    clust = OPTICS(41        min_samples=int(min_samples),42        xi=float(xi),43        min_cluster_size=float(min_cluster_size),44    )45 46    # Run the fit47    clust.fit(X)48 49    labels_050 = cluster_optics_dbscan(50        reachability=clust.reachability_,51        core_distances=clust.core_distances_,52        ordering=clust.ordering_,53        eps=0.5,54    )55    labels_200 = cluster_optics_dbscan(56        reachability=clust.reachability_,57        core_distances=clust.core_distances_,58        ordering=clust.ordering_,59        eps=2,60    )61 62    space = np.arange(len(X))63    reachability = clust.reachability_[clust.ordering_]64    labels = clust.labels_[clust.ordering_]65 66    plt.figure(figsize=(10, 6))67    G = gridspec.GridSpec(2, 3)68    ax1 = plt.subplot(G[0, :])69    ax2 = plt.subplot(G[1, 0])70    ax3 = plt.subplot(G[1, 1])71    ax4 = plt.subplot(G[1, 2])72 73    # Reachability plot74    colors = ["g.", "r.", "b.", "y.", "c."]75    for klass, color in zip(range(0, 5), colors):76        Xk = space[labels == klass]77        Rk = reachability[labels == klass]78        ax1.plot(Xk, Rk, color, alpha=0.3)79    ax1.plot(space[labels == -1], reachability[labels == -1], "k.", alpha=0.3)80    ax1.plot(space, np.full_like(space, 2.0, dtype=float), "k-", alpha=0.5)81    ax1.plot(space, np.full_like(space, 0.5, dtype=float), "k-.", alpha=0.5)82    ax1.set_ylabel("Reachability (epsilon distance)")83    ax1.set_title("Reachability Plot")84 85    # OPTICS86    colors = ["g.", "r.", "b.", "y.", "c."]87    for klass, color in zip(range(0, 5), colors):88        Xk = X[clust.labels_ == klass]89        ax2.plot(Xk[:, 0], Xk[:, 1], color, alpha=0.3)90    ax2.plot(X[clust.labels_ == -1, 0], X[clust.labels_ == -1, 1], "k+", alpha=0.1)91    ax2.set_title("Automatic Clustering\nOPTICS")92 93    # DBSCAN at 0.594    colors = ["g.", "r.", "b.", "c."]95    for klass, color in zip(range(0, 4), colors):96        Xk = X[labels_050 == klass]97        ax3.plot(Xk[:, 0], Xk[:, 1], color, alpha=0.3)98    ax3.plot(X[labels_050 == -1, 0], X[labels_050 == -1, 1], "k+", alpha=0.1)99    ax3.set_title("Clustering at 0.5 epsilon cut\nDBSCAN")100 101    # DBSCAN at 2.102    colors = ["g.", "m.", "y.", "c."]103    for klass, color in zip(range(0, 4), colors):104        Xk = X[labels_200 == klass]105        ax4.plot(Xk[:, 0], Xk[:, 1], color, alpha=0.3)106    ax4.plot(X[labels_200 == -1, 0], X[labels_200 == -1, 1], "k+", alpha=0.1)107    ax4.set_title("Clustering at 2.0 epsilon cut\nDBSCAN")108 109    plt.tight_layout()110 111    return plt112 113 114title = "Demo of OPTICS clustering algorithm"115with gr.Blocks(title=title, theme=theme) as demo:116    gr.Markdown(f"## {title}")117    gr.Markdown(118        "[Scikit-learn Example](https://scikit-learn.org/stable/auto_examples/cluster/plot_optics.html)"119    )120 121    gr.Markdown(122        "Finds core samples of high density and expands clusters from them. This example uses data that is \123        generated so that the clusters have different densities. The [OPTICS](https://scikit-learn.org/stable/modules/generated/sklearn.cluster.OPTICS.html#sklearn.cluster.OPTICS) is first used with its Xi cluster detection \124        method, and then setting specific thresholds on the reachability, which corresponds to [DBSCAN](https://scikit-learn.org/stable/modules/generated/sklearn.cluster.DBSCAN.html#sklearn.cluster.DBSCAN). We can see that \125        the different clusters of OPTICS’s Xi method can be recovered with different choices of thresholds in DBSCAN."126    )127 128    with gr.Row().style(equal_height=True):129        with gr.Column(scale=0.75):130            n_points_per_cluster = gr.Slider(131                minimum=200,132                maximum=500,133                label="Number of points per cluster",134                step=50,135                value=250,136            )137            with gr.Row(visible=False):138                gr.Markdown("##")139 140            min_samples = gr.Slider(141                minimum=10,142                maximum=100,143                label="OPTICS - Minimum number of samples",144                step=5,145                value=50,146                info="The number of samples in a neighborhood for a point to be considered as a core point.",147            )148            with gr.Row(visible=False):149                gr.Markdown("##")150 151            xi = gr.Slider(152                minimum=0,153                maximum=0.2,154                label="OPTICS - Xi",155                step=0.01,156                value=0.05,157                info="Determines the minimum steepness on the reachability plot that constitutes a cluster boundary. ",158            )159            with gr.Row(visible=False):160                gr.Markdown("##")161            min_cluster_size = gr.Slider(162                minimum=0.01,163                maximum=0.1,164                label="OPTICS - Minimum cluster size",165                step=0.01,166                value=0.05,167                info="Minimum number of samples in an OPTICS cluster, expressed as an absolute number or a fraction of the number of samples (rounded to be at least 2).",168            )169 170        plt_out = gr.Plot()171 172        n_points_per_cluster.change(173            do_submit,174            inputs=[n_points_per_cluster, min_samples, xi, min_cluster_size],175            outputs=plt_out,176        )177        min_samples.change(178            do_submit,179            inputs=[n_points_per_cluster, min_samples, xi, min_cluster_size],180            outputs=plt_out,181        )182        xi.change(183            do_submit,184            inputs=[n_points_per_cluster, min_samples, xi, min_cluster_size],185            outputs=plt_out,186        )187        min_cluster_size.change(188            do_submit,189            inputs=[n_points_per_cluster, min_samples, xi, min_cluster_size],190            outputs=plt_out,191        )192 193 194 195if __name__ == "__main__":196    demo.launch()197 198