Compare LogisticRegression solver with sklearn’s liblinear backend

file_sizes:   0%|                                   | 0.00/26.8M [00:00<?, ?B/s]
file_sizes:   0%|                           | 24.6k/26.8M [00:00<03:50, 116kB/s]
file_sizes:   0%|                           | 49.2k/26.8M [00:00<03:52, 115kB/s]
file_sizes:   0%|                            | 106k/26.8M [00:00<02:24, 185kB/s]
file_sizes:   1%|▏                           | 221k/26.8M [00:00<01:22, 323kB/s]
file_sizes:   2%|▍                           | 451k/26.8M [00:01<00:44, 593kB/s]
file_sizes:   3%|▉                          | 909k/26.8M [00:01<00:23, 1.12MB/s]
file_sizes:   7%|█▊                        | 1.83M/26.8M [00:01<00:11, 2.15MB/s]
file_sizes:  14%|███▌                      | 3.66M/26.8M [00:01<00:05, 4.19MB/s]
file_sizes:  27%|███████                   | 7.33M/26.8M [00:01<00:02, 8.22MB/s]
file_sizes:  39%|██████████▏               | 10.5M/26.8M [00:02<00:01, 10.1MB/s]
file_sizes:  57%|██████████████▊           | 15.2M/26.8M [00:02<00:00, 13.3MB/s]
file_sizes:  70%|██████████████████▎       | 18.9M/26.8M [00:02<00:00, 14.3MB/s]
file_sizes:  84%|█████████████████████▉    | 22.5M/26.8M [00:02<00:00, 14.9MB/s]
file_sizes:  98%|█████████████████████████▍| 26.2M/26.8M [00:03<00:00, 15.3MB/s]
file_sizes: 100%|██████████████████████████| 26.8M/26.8M [00:03<00:00, 8.74MB/s]

import time
import warnings
import numpy as np
from numpy.linalg import norm
import matplotlib.pyplot as plt
from sklearn import linear_model
from libsvmdata import fetch_libsvm

from celer import LogisticRegression

warnings.filterwarnings("ignore", message="Objective did not converge")
warnings.filterwarnings("ignore", message="Liblinear failed to converge")

X, y = fetch_libsvm("news20.binary")

C_min = 2 / norm(X.T @ y, ord=np.inf)
C = 20 * C_min


def pobj_logreg(w):
    return np.sum(np.log(1 + np.exp(-y * (X @ w)))) + 1. / C * norm(w, ord=1)


pobj_celer = []
t_celer = []

for n_iter in range(10):
    t0 = time.time()
    clf = LogisticRegression(
        C=C, solver="celer-pn", max_iter=n_iter, tol=0).fit(X, y)
    t_celer.append(time.time() - t0)
    w_celer = clf.coef_.ravel()
    pobj_celer.append(pobj_logreg(w_celer))

pobj_celer = np.array(pobj_celer)


pobj_libl = []
t_libl = []

for n_iter in np.arange(0, 50, 10):
    t0 = time.time()
    clf = linear_model.LogisticRegression(
        C=C, solver="liblinear", penalty='l1', fit_intercept=False,
        max_iter=n_iter, random_state=0, tol=1e-10).fit(X, y)
    t_libl.append(time.time() - t0)
    w_libl = clf.coef_.ravel()
    pobj_libl.append(pobj_logreg(w_libl))

pobj_libl = np.array(pobj_libl)

p_star = min(pobj_celer.min(), pobj_libl.min())

plt.close("all")
fig = plt.figure(figsize=(4, 2), constrained_layout=True)
plt.semilogy(t_celer, pobj_celer - p_star, label="Celer-PN")
plt.semilogy(t_libl, pobj_libl - p_star, label="liblinear")
plt.legend()
plt.xlabel("Time (s)")
plt.ylabel("objective suboptimality")
plt.show(block=False)