BiArchetype Analysis

from archetypes.datasets import make_archetypal_dataset
import numpy as np


n_archetypes = (3, 3)

rng = np.random.default_rng(5)
archetypes = rng.uniform(size=n_archetypes)

import matplotlib.pyplot as plt

plt.figure(figsize=(5, 5))
plt.imshow(archetypes, cmap="Blues")
plt.show()

X, labels = make_archetypal_dataset(
    shape=(100, 100), alpha=0.4, archetypes=archetypes, noise=0.01, generator=123
)

import matplotlib.pyplot as plt

plt.figure(figsize=(5, 5))
plt.imshow(X, cmap="Blues")
plt.show()

from archetypes.jax import BiAA

method_kwargs = {"optimizer": "sgd", "optimizer_kwargs": {"learning_rate": 5e-2}}

model = BiAA(
    n_archetypes=n_archetypes,
    max_iter=1_000,
    tol=1e-3,
    method="autogd",
    method_kwargs=method_kwargs,
    init="uniform",
    seed=123,
)

model.fit(X)

BiAA(init_kwargs={}, max_iter=1000,
     method_kwargs={'optimizer': 'sgd',
                    'optimizer_kwargs': {'learning_rate': 0.05}},
     n_archetypes=(3, 3), seed=123, tol=0.001)

In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.

plt.figure(figsize=(5, 5))
plt.imshow(model.archetypes_, cmap="Blues")
plt.show()

import matplotlib.pyplot as plt
from archetypes.visualization import simplex

plt.figure(figsize=(10, 10))
simplex(
    model.similarity_degree_[0],
    c=labels[0],
    alpha=0.5,
    show_circle=False,
    show_direction=True,
)

plt.show()

plt.plot(model.loss_)

[<matplotlib.lines.Line2D at 0x163530a7e20>]