How to Fix Missing 3D Plots in Python Logistic Regression Visualizations

1. Introduction

I'm a Python enthusiast. Recently a member asked why a 3‑D plot generated with Matplotlib does not appear. Below is the original code that attempts to plot a logistic‑regression loss surface.

from sklearn.datasets import load_breast_cancer
from sklearn.linear_model import LogisticRegression
import numpy as np
from matplotlib import pyplot as plt
from mpl_toolkits.mplot3d import Axes3D

# Load data
data = load_breast_cancer()
X, y = data['data'][:, :2], data['target']

# Fit logistic regression without intercept
lr = LogisticRegression(fit_intercept=False)
lr.fit(X, y)
theta1 = lr.coef_[0, 0]
theta2 = lr.coef_[0, 1]
print(theta1)
print(theta2)

def p_theta_function(features, w1, w2):
    """Compute the predicted probability given w1 and w2."""
    z = w1 * features[0] + w2 * features[1]
    return 1 / (1 + np.exp(-z))

def loss_function(sample_features, sample_labels, w1, w2):
    """Calculate the total loss for a set of samples."""
    result = 0
    for feature, label in zip(sample_features, sample_labels):
        p = p_theta_function(feature, w1, w2)
        loss = -label * np.log(p) - (1 - label) * np.log(1 - p)
        result += loss
    return result

theta1_space = np.linspace(theta1-0.6, theta1+0.6, 50)
theta2_space = np.linspace(theta2-0.6, theta2+0.6, 50)
result1 = np.array([loss_function(X, y, i, theta2) for i in theta1_space])
result2 = np.array([loss_function(X, y, theta1, i) for i in theta2_space])

fig1 = plt.figure(figsize=(8, 6))
plt.subplot(221)
plt.plot(theta1_space, result1)

plt.subplot(222)
plt.plot(theta2_space, result2)

plt.subplot(223)
theta1_grid, theta2_grid = np.meshgrid(theta1_space, theta2_space)
loss_grid = loss_function(X, y, theta1_grid, theta2_grid)
plt.contour(theta1_grid, theta2_grid, loss_grid)

plt.subplot(224)
plt.contour(theta1_grid, theta2_grid, loss_grid, 30)

fig2 = plt.figure()
ax = Axes3D(fig2)
ax.plot_surface(theta1_grid, theta2_grid, loss_grid)

plt.show()

2. Implementation Details

A collaborator suggested trying the code on Windows. Another user ran it successfully on a machine with Matplotlib 3.5, producing the plots shown below.

The two figures illustrate that the 2‑D loss curves render correctly, while the 3‑D surface may fail on low‑performance machines or with newer Matplotlib versions (e.g., 3.6+ where Axes3D(fig) is deprecated). Using Matplotlib 3.5 avoids the warning and produces the expected 3‑D plot.

3. Conclusion

This article demonstrates how to compute and visualize the logistic‑regression loss surface in Python, highlights version‑specific issues with Matplotlib’s 3‑D API, and provides a working example that runs on Matplotlib 3.5.