Exploring Optuna
Optuna is a hyperparameter tuning package in Python that provides several hyperparameter tuning methods. We discussed how to utilize Optuna to conduct a hyperparameter tuning experiment in Chapter 9, Hyperparameter Tuning via Optuna. Here, we will discuss how to utilize this package to track those experiments.
Similar to Scikit-Optimize, Optuna provides very nice visualization modules to help us track the hyperparameter tuning experiments and as a guide for us to decide which subspace to search in the next trial. Four visualization modules can be utilized, as shown here. All of them expect the study object (see Chapter 9, Hyperparameter Tuning via Optuna) as input. Please see the full code in this book’s GitHub repository:
plot_contour: This is used to visualize the relationship between hyperparameters (as well as the objective function scores) in the form of contour plots:
Figure 13.10 – Contour plot
plot_optimization_history...
