Generating Multiple Similar Datasets

Using repliclust, you can generate many different synthetic data sets that all look similar. To illustrate, we will now generate nine different data sets based on the “oblong” archetype defined in the Basic Usage section. In the simulation below, we define the archetype and data generator again for convenience.

import matplotlib.pyplot as plt
from repliclust import Archetype, DataGenerator, set_seed

set_seed(1)

archetype_oblong = Archetype(n_clusters=5, dim=2, n_samples=500,
                            aspect_ref=3, aspect_maxmin=1.5,
                            name="oblong")
data_generator = DataGenerator(archetype=archetype_oblong)

fig, ax = plt.subplots(figsize=(9,9), dpi=300, nrows=3, ncols=3)

for i in range(3):
    for j in range(3):
        X, y, archetype = data_generator.synthesize(quiet=True)
        ax[i,j].set_title('Dataset #' + str(i*3 + (j+1)), fontsize=10)
        ax[i,j].scatter(X[:,0],X[:,1],c=y, s=5, alpha=0.5, linewidth=0.3)
        ax[i,j].set_xticks([]); ax[i,j].set_yticks([])

plt.subplots_adjust(hspace=0.20)
fig.suptitle("Synthetic Data from Archetype '"
            + archetype.name + "'", y=0.97)

Setting the option quiet=True in the call to synthesize() avoids printing status updates during data generation.