The objective range of Matplotlib contour plot has a different numerical range from Plotly backend

Expected behavior

The range of objective in optuna.visualization.matplotlib.plot_contour should have a similar range to optuna.visualization.plot_contour.

Environment

• Optuna version: 2.7.0
• Python version: 3.8.8
• OS: Mac
• (Optional) Other libraries and their versions:
  • matplotlib: 3.3.4
  • Plotly: 4.11.0

Error messages, stack traces, or logs

The matplotlib range is from -60 000 to 360 000 as in the following figure.

On the other hand, Plotly’s range is from 0 to 9 000. That is supposed to be true.

Steps to reproduce

1. Please run a following example code blocks

Reproducible examples (optional)

import optuna


def objective(trial):
    x = trial.suggest_uniform('x', -100, 100)
    y = trial.suggest_categorical('y', list(range(-10, 10)))
    z = trial.suggest_uniform('z', -100, 100)
    return x ** 2 + y - z

study = optuna.create_study(sampler=optuna.samplers.TPESampler(seed=7))
study.optimize(objective, n_trials=200)

optuna.visualization.matplotlib.plot_contour(study, params=['y', 'z'])
optuna.visualization.plot_contour(study, params=['y', 'z']) # plotly

[nzw0301]

Brilliant! Thank you for sharing your investigation. It sounds great – reviewers can understand the change in the pull request smoothly.

[01-vyom]

I was reading the provided link for interpolation of irregular space data and tried the CubicInterpolation code [interpolating z], still got the same results (incorrect numeric range). The Linear interpolation and the option linear in the griddata function works in the same way, so we can just use that if we go for linear interpolation. Let me know your thoughts…