T-SNE visualization by `sklearn.manifold.TSNE`

draw_tsne.py

import numpy as np
from sklearn.manifold import TSNE
from matplotlib import pyplot as plt


def draw_tsne(X, y, save_fig=None, show=False):
    """
    T-SNE visualization by `sklearn.manifold.TSNE`.
    Reference: https://www.scipy-lectures.org/packages/scikit-learn/auto_examples/plot_tsne.html
    :param X: data to be projected
    :param y: data labels
    :param save_fig: the path to save the fig
    :return: None
    """
    assert len(X) == len(y)

    t_sne = TSNE(n_components=2, random_state=0, verbose=1)
    X_2d = t_sne.fit_transform(X)

    num_ids = len(np.unique(y))
    target_ids = range(num_ids)

    plt.figure(figsize=(6, 5))
    # distinct color map, https://stackoverflow.com/questions/14720331/how-to-generate-random-colors-in-matplotlib
    cmap = plt.cm.get_cmap('hsv', num_ids+1)
    for i in target_ids:
        plt.scatter(X_2d[y == i, 0], X_2d[y == i, 1], c=cmap(i), s=2, label=str(i))
    plt.legend()
    if save_fig:
        plt.savefig(save_fig)
    if show:
        plt.show()