line_models_sampling.py

import numpy as np
import pandas as pd

def compute_line_parameters(slope, intercept, mean, std, num_points, x_min, x_max, num_iterations):
    line_parameters = np.zeros((num_iterations, 2))
    for i in range(num_iterations):
        # Generate points on the line with noise
        x_values, y_values_with_noise = predict_line_with_noise(slope, intercept, mean, std, num_points, x_min, x_max)

        # Randomly sample two points from the line
        idx = np.random.choice(num_points, 2, replace=False)
        x1, y1 = x_values[idx[0]], y_values_with_noise[idx[0]]
        x2, y2 = x_values[idx[1]], y_values_with_noise[idx[1]]

        # Compute slope and intercept of line passing through sampled points
        line_slope = (y2 - y1) / (x2 - x1)
        line_intercept = y1 - line_slope * x1

        line_parameters[i, 0] = line_slope
        line_parameters[i, 1] = line_intercept

    return line_parameters

from sklearn.cluster import KMeans
# import matplotlib.pyplot as plt

import gradio as gr
import matplotlib.pyplot as plt
import numpy as np
from sklearn.cluster import KMeans

def cluster_line_parameters(line_parameters, num_clusters):
    # Cluster line parameters using KMeans
    kmeans = KMeans(n_clusters=num_clusters)
    kmeans.fit(line_parameters)
    labels = kmeans.labels_
    centroids = kmeans.cluster_centers_
    # how many points are in each cluster
    counts = np.bincount(labels)

    # Create a scatter plot of the data points
    fig, ax = plt.subplots()
    ax.scatter(line_parameters[:, 0], line_parameters[:, 1], c=labels, cmap='viridis')
    ax.set_xlabel('Slope')
    ax.set_ylabel('Intercept')

    # Generate some lines using the centroids
    '''
    x = np.linspace(x_min, x_max, num_points)
    for i in range(num_clusters):
        slope, intercept = centroids[i]
        y = slope * x + intercept
        ax.plot(x, y, linewidth=2)
    '''

    # Convert the figure to a PNG image
    fig.canvas.draw()
    img = np.frombuffer(fig.canvas.tostring_rgb(), dtype=np.uint8)
    img = img.reshape(fig.canvas.get_width_height()[::-1] + (3,))

    # Close the figure to free up memory
    plt.close(fig)

    # Return the labels, centroids, and image
    return labels, centroids, counts, img

import gradio as gr

def predict_line_with_noise(slope, intercept, mean, std, num_points, x_min, x_max):
    x_values = np.linspace(x_min, x_max, num_points)
    noise = np.random.normal(mean, std, num_points)
    y_values = slope * x_values + intercept + noise
    return x_values, y_values

def cluster_line_params(slope, intercept, mean, std, num_points, x_min, x_max, num_iterations, num_clusters):
    num_points = int(num_points)
    num_iterations = int(num_iterations)
    num_clusters = int(num_clusters)
    line_parameters = compute_line_parameters(slope, intercept, mean, std, num_points, x_min, x_max, num_iterations)
    labels, centroids, counts, img = cluster_line_parameters(line_parameters, num_clusters)
    # return labels, centroids, img
    # put counts as the third column of centroids
    centroids = np.c_[centroids, counts]
    df = pd.DataFrame(centroids, columns=["Slope", "Intercept", "Count"])
    # return img, centroids, df
    return img, df

# Define input and output components
inputs = [
    gr.inputs.Slider(minimum=-5.0, maximum=5.0, default=1.0, label="Slope"),
    gr.inputs.Slider(minimum=-10.0, maximum=10.0, default=0.0, label="Intercept"),
    gr.inputs.Slider(minimum=0.0, maximum=5.0, default=1.0, label="Mean"),
    gr.inputs.Slider(minimum=0.0, maximum=5.0, default=1.0, label="Standard Deviation"),
    gr.inputs.Number(default=100, label="Number of Points"),
    gr.inputs.Number(default=-5, label="Minimum Value of x"),
    gr.inputs.Number(default=5, label="Maximum Value of x"),
    gr.inputs.Number(default=100, label="Number of Iterations"),
    gr.inputs.Number(default=3, label="Number of Clusters")
]

outputs = [
    #image and numpy
    gr.outputs.Image(label="Image", type="pil"),
    # show the centroids and counts, which is a numpy array
    gr.outputs.Dataframe(label="Centroids and Counts", type="pandas")
]

# Create the Gradio interface
interface = gr.Interface(
    fn=cluster_line_params,
    inputs=inputs,
    # outputs=["numpy", "numpy", "image"],
    outputs=outputs,
    title="Line Parameter Clustering",
    description="Cluster line parameters with Gaussian noise",
    allow_flagging=False
)

# Launch the interface
interface.launch(share=True)