TOKYO – September 2, 2024 – Preferred Networks, Inc. (PFN) today released Optuna™ v4.0, the fourth major update of the open-source hyperparameter optimization framework for machine learning, first initiated by PFN.

Optuna now has over 10,000 stars on GitHub and is used in over 16,000 software applications, making it one of the most popular hyperparameter optimization framework. With the release of v4.0, PFN developed the feature-sharing platform OptunaHub, enhanced experiment management features, and improved support for distributed optimization.

Official Release of Feature-Sharing Platform OptunaHub

PFN’s Optuna development team has officially released OptunaHub, a feature-sharing platform for Optuna. State-of-the-art optimization algorithms and domain-specific methods are currently available in OptunaHub. Contributors can easily register their methods and deliver them to Optuna users around the world.

Enhancement of Experiment Management Features

The management mechanism for files generated during hyperparameter optimization experiments is now officially supported. This enables users to manage large-sized data, such as text, image, and audio files produced by generative AI and trained machine learning models within the Optuna ecosystem.

Enhanced support for distributed optimization

A new distributed optimization feature using Network File System is now officially supported. This makes it easy to perform distributed optimization using multiple computers, even in environments where it is difficult to build a database server such as on supercomputers.

In v4.0, PFN has also added and improved other Optuna features and algorithms. For more details, please read the release blog.

About Optuna

Optuna is a hyperparameter optimization framework written in Python. When combined with various machine learning software tools, Optuna can automate trial-and-error processes for hyperparameter search and helps control algorithm behaviors and increase accuracy. Since open-sourced by PFN in December 2018, many external contributors have joined in its development. Optuna has been used for a variety of products and projects, including performance enhancement of Preferred Robotics’s autonomous mobile robot Kachaka and crystal structure prediction using the atomistic simulator Matlantis™ for materials discovery.

• Website https://optuna.org
• Document https://optuna.readthedocs.io/en/stable/
• Tutorial https://optuna.readthedocs.io/en/stable/tutorial/
• GitHub https://github.com/optuna/optuna
• OptunaHub https://hub.optuna.org