This paper introduces Optuna, a next-generation hyperparameter optimization framework designed to address the limitations of existing frameworks. The key design criteria for Optuna include: 1. **Define-by-run API**: Allows users to dynamically construct the parameter search space. 2. **Efficient Implementation of Search and Pruning Strategies**: Combines advanced sampling and pruning algorithms to enhance performance. 3. **Versatile and Easy-to-Setup Architecture**: Supports various tasks, from lightweight experiments to distributed computing. Optuna's *define-by-run* principle enables users to dynamically construct the search space, making it highly modular and flexible. The framework supports both relational and independent sampling methods, and includes an efficient pruning algorithm to terminate unpromising trials early. Optuna is designed to be scalable and versatile, suitable for a wide range of applications, from lightweight interactive interfaces to heavy-duty distributed computations. The paper presents experimental results demonstrating Optuna's superior performance compared to other frameworks in terms of both efficiency and effectiveness. Optuna has been successfully applied in various real-world projects, including machine learning competitions, high-performance computing, and multimedia frameworks. The authors conclude that the new design criteria for next-generation optimization frameworks are valuable and hope that Optuna will serve as a foundation for future developments.This paper introduces Optuna, a next-generation hyperparameter optimization framework designed to address the limitations of existing frameworks. The key design criteria for Optuna include: 1. **Define-by-run API**: Allows users to dynamically construct the parameter search space. 2. **Efficient Implementation of Search and Pruning Strategies**: Combines advanced sampling and pruning algorithms to enhance performance. 3. **Versatile and Easy-to-Setup Architecture**: Supports various tasks, from lightweight experiments to distributed computing. Optuna's *define-by-run* principle enables users to dynamically construct the search space, making it highly modular and flexible. The framework supports both relational and independent sampling methods, and includes an efficient pruning algorithm to terminate unpromising trials early. Optuna is designed to be scalable and versatile, suitable for a wide range of applications, from lightweight interactive interfaces to heavy-duty distributed computations. The paper presents experimental results demonstrating Optuna's superior performance compared to other frameworks in terms of both efficiency and effectiveness. Optuna has been successfully applied in various real-world projects, including machine learning competitions, high-performance computing, and multimedia frameworks. The authors conclude that the new design criteria for next-generation optimization frameworks are valuable and hope that Optuna will serve as a foundation for future developments.