app.py

"""
Demo is based on https://scikit-learn.org/stable/auto_examples/feature_selection/plot_rfe_digits.html
"""
from sklearn.svm import SVC
from sklearn.datasets import load_digits
from sklearn.feature_selection import RFE
import matplotlib.pyplot as plt

# Load the digits dataset
digits = load_digits()
X = digits.images.reshape((len(digits.images), -1))
y = digits.target

# Create the RFE object and rank each pixel
svc = SVC(kernel="linear", C=1)


def recursive_feature_elimination(n_features_to_select, step, esimator=svc):
    # Plot the results
    fig = plt.figure()
    rfe = RFE(estimator=esimator, n_features_to_select=1, step=1)
    # step : Number of feature to remove at each iteration, least important are removed
    # n_features_to_select : Number of features to be selected after repeated elimination
    rfe.fit(X, y)
    ranking = rfe.ranking_.reshape(digits.images[0].shape)

    # Plot pixel ranking
    plt.matshow(ranking, cmap=plt.cm.Blues)
    plt.colorbar()
    plt.title("Ranking of pixels with RFE")
    # plt.show()
    return plt


import gradio as gr

title = " Illustration of Recursive feature elimination.🌲 "

with gr.Blocks(title=title) as demo:
    gr.Markdown(f"# {title}")
    gr.Markdown(
        "This example demonstrates recursive feature elimination. <br>"
        "Dataset  is `load_digits()` which is images of size 8x8 images of hand-written digits.  <br>"
        "**Parameters** <br> **Number of features to select**: Represents the features left at the end of feature selection process. <br>"
        "**Step**: Number of feature to remove at each iteration, least important are removed. <br>"
    )

    gr.Markdown(
        "Support vector classifier is used as estimator to rank features. <br>"
    )

    gr.Markdown(
        "Demo is based on [sklearn docs](https://scikit-learn.org/stable/auto_examples/feature_selection/plot_rfe_digits.html)."
    )
    with gr.Row():
        n_features_to_select = gr.Slider(
            minimum=0, maximum=20, step=1, value=1, label="Number of features to select"
        )
        step = gr.Slider(minimum=0, maximum=20, step=1, value=1, label="Step")

    btn = gr.Button(value="Submit")

    btn.click(
        recursive_feature_elimination,
        inputs=[n_features_to_select, step],
        outputs=gr.Plot(
            label="Recursive feature elimination of pixels in digit classification"
        ),
    )  #

    gr.Markdown(
        "Plot shows the importance of each pixel in the classification of the digits. <br>"
    )

demo.launch()