app.py

"""
=======================================================
Comparison of LDA and PCA 2D projection of Iris dataset
=======================================================

The Iris dataset represents 3 kind of Iris flowers (Setosa, Versicolour
and Virginica) with 4 attributes: sepal length, sepal width, petal length
and petal width.

Principal Component Analysis (PCA) applied to this data identifies the
combination of attributes (principal components, or directions in the
feature space) that account for the most variance in the data. Here we
plot the different samples on the 2 first principal components.

Linear Discriminant Analysis (LDA) tries to identify attributes that
account for the most variance *between classes*. In particular,
LDA, in contrast to PCA, is a supervised method, using known class labels.

"""

import matplotlib.pyplot as plt
import gradio as gr
from sklearn import datasets
from sklearn.decomposition import PCA
from sklearn.discriminant_analysis import LinearDiscriminantAnalysis

# load data
iris = datasets.load_iris()

X = iris.data
y = iris.target
target_names = iris.target_names

def plot_lda_pca(n_samples = 100,
                 n_components=2,
                 n_features=4):
    
    '''
    Function to plot LDA and PCA clustering.
    
    Parameters
    ----------
    n_components : int, default=2
        Number of components to keep.

    n_features : int, default=5
        Number of features to generate.
    
    Returns
    -------
    fig : matplotlib.pyplot.figure
        Figure object.
    '''
    
    
    
    # take sample of data
    X = X[:n_samples, :n_features]
    y = y[:n_samples]
    
    # fit PCA
    pca = PCA(n_components=n_components)
    X_r = pca.fit(X).transform(X)
    print(f"shape of X_r: {X_r.shape}")
    # fit LDA
    lda = LinearDiscriminantAnalysis(n_components=n_components)
    X_r2 = lda.fit(X, y).transform(X)
    print(f"shape of X_r2: {X_r2.shape}")
    # take first two components
    X_r = X_r[:, :2]
    X_r2 = X_r2[:, :2]
    
    print(f"shape of X_r after: {X_r.shape}")
    print(f"shape of X_r2 after: {X_r2.shape}")
    
    # Percentage of variance explained for each components
    print(
        "explained variance ratio (first two components): %s"
        % str(pca.explained_variance_ratio_)
    )

    # fig = plt.figure(1, facecolor="w", figsize=(5,5))
    fig, axes = plt.subplots(2,1, sharey= False, sharex=False, figsize = (8,6))
    colors = ["navy", "turquoise", "darkorange"]
    lw = 2

    for color, i, target_name in zip(colors, [0, 1, 2], target_names):
        axes[0].scatter(
            X_r[y == i, 0], X_r[y == i, 1], color=color, alpha=0.8, lw=lw, label=target_name
        )
    axes[0].legend(loc="lower right")
    axes[0].set_title("PCA of IRIS dataset")    
    for color, i, target_name in zip(colors, [0, 1, 2], target_names):
        axes[1].scatter(
            X_r2[y == i, 0], X_r2[y == i, 1], alpha=0.8, color=color, label=target_name
        )
    plt.legend(loc="best", shadow=False, scatterpoints=1)
    axes[1].legend(loc="lower right")
    axes[1].set_title("LDA of IRIS dataset")
    plt.tight_layout()   


    return fig


title = "2-D projection of Iris dataset using LDA and PCA"
with gr.Blocks(title=title) as demo:
    gr.Markdown(f"# {title}")
    gr.Markdown(" This example shows how one can use Prinicipal Components Analysis (PCA) and Factor Analysis (FA) for model selection by observing the likelihood of a held-out dataset with added noise <br>"
    " The number of samples (n_samples) will determine the number of data points to produce.  <br>"
    " The number of components (n_components) will determine the number of components each method will fit to, and will affect the likelihood of the held-out set.  <br>"
    " The number of features (n_components) determine the number of features the toy dataset X variable will have.  <br>"
    " For further details please see the sklearn docs:"    
    )

    gr.Markdown(" **[Demo is based on sklearn docs found here](https://scikit-learn.org/stable/auto_examples/decomposition/plot_pca_vs_lda.html#sphx-glr-auto-examples-decomposition-plot-pca-vs-lda-py)** <br>")

    gr.Markdown(" **Dataset** : A toy dataset with corrupted with homoscedastic noise (noise variance is the same for each feature) or heteroscedastic noise (noise variance is the different for each feature) . <br>")
    gr.Markdown(" Different number of features and number of components affect how well the low rank space is recovered. <br>"
                "  Larger Depth trying to overfit and learn even the finner details of the data.<br>"
               )
    # set max samples
    max_samples = len(iris.data)
    with gr.Row():        
        n_samples = gr.Slider(value=100, minimum=2, maximum=max_samples, step=1, label="n_samples")
        n_features = gr.Slider(value=4, minimum=2, maximum=4, step=1, label="n_features")
        

    btn = gr.Button(value="Run")
    btn.click(plot_lda_pca,inputs= [n_samples, n_features], outputs= gr.Plot(label='PCA vs LDA clustering') ) # 
    

demo.launch()