Calf runtime complexity

A classifier runtime comparison plot including Calf. Lasso is much faster than LogisticRegression, which in turn is faster than Calf. The calculation of AUC dominates the Calf runtime.

# Author: Rolf Carlson, Carlson Research LLC, <hrolfrc@gmail.com>
# License: 3-clause BSD

import time

import matplotlib.pyplot as plt
import numpy as np
from sklearn.datasets import make_classification
from sklearn.linear_model import Lasso, LogisticRegression

from calfcv import Calf

ts = {'calf': [], 'lr': [], 'las': []}
ns = np.arange(1, 10, 1)
for i in ns:
    X, y = make_classification(
        n_samples=100 * i,
        n_features=20 * i,
        n_informative=10 * i,
        n_redundant=5 * i,
        n_classes=2,
        random_state=11
    )

    ts['calf'].append(Calf().fit(X, y).fit_time_)

    start = time.time()
    cls1 = LogisticRegression(max_iter=10000).fit(X, y)
    elapsed = time.time() - start
    ts['lr'].append(elapsed)

    start = time.time()
    cls2 = Lasso(max_iter=10000).fit(X, y)
    elapsed = time.time() - start
    ts['las'].append(elapsed)

degree = 1

# stack vertically, share the same x-axis
# figure = plt.figure(figsize=(10, 10))
fig, ax = plt.subplots(5, sharex=True)
fig.suptitle('Calf fit times versus LogisticRegression and Lasso ')

coeffs = np.polyfit(ns, ts['calf'], degree)
p = np.poly1d(coeffs)
ax[0].plot(ns, ts['calf'], 'or')
ax[0].plot(ns, [p(n) for n in ns], '-g')
ax[0].set_ylabel('seconds')
ax[0].set_title('Calf fit times')

coeffs = np.polyfit(ns, ts['lr'], degree)
p = np.poly1d(coeffs)
ax[1].plot(ns, ts['lr'], 'or')
ax[1].plot(ns, [p(n) for n in ns], '-b')
ax[1].set_ylabel('seconds')
ax[1].set_title('LogisticRegression fit times')

ax[2].plot(ns, np.divide(ts['calf'], ts['lr']), 'or')
ax[2].set_title('Speedup of LogisticRegression over Calf')
ax[2].set_ylabel('speedup')

coeffs = np.polyfit(ns, ts['las'], degree)
p = np.poly1d(coeffs)
ax[3].plot(ns, ts['las'], 'or')
ax[3].plot(ns, [p(n) for n in ns], '-b')
ax[3].set_ylabel('seconds')
ax[3].set_title('Lasso fit times')

ax[4].plot(ns, np.divide(ts['calf'], ts['las']), 'or')
ax[4].set_title('Speedup of Lasso over Calf')
ax[4].set_ylabel('speedup')

plt.tight_layout()
plt.show()