OptunaAutoML

Optuna v4.6 has been released! 📊Optuna Dashboard LLM Integration 🚀Significant speedup of GPSampler 🦾Full support for multi-objective and constrained optimization in AutoSampler ⛏️Robust Bayesian Optimization Algorithms medium.com/optuna/announc…

Optuna v4.8.0 has been released! 🚀 This release introduces the constant liar strategy to GPSampler for improved distributed optimization (currently for single-objective & unconstrained tasks). Further improvements are coming in v4.9.0! 👇See what’s new github.com/optuna/optuna/…

Optuna v4.7.0 has been released! This is a maintenance release with various minor bug fixes and improvements to the documentation and more. 👇See what’s new github.com/optuna/optuna/…

Nice work! This research uses Optuna for multi-objective black-box optimization in manipulator design.

Honored to be interviewed about the history of Optuna. It covers everything from our early challenges in 2018 to the mindset I carry into my work today. Full interview (in Japanese): findy-code.io/engineer-lab/i…

Optuna Dashboard now supports LLM integration🎉 You can now ask in natural language to generate Plotly charts or filter trials. Learn how we designed and implemented these LLM features. medium.com/optuna/optuna-…

AutoSampler now fully supports multi-objective & constrained optimization! AutoSampler lets you efficiently solve a wide range of problems without worrying about which optimization algorithm to use. Learn more in our article: medium.com/optuna/autosam…

Our Optuna paper has reached over 10,000 citations! A huge thank you to the amazing co-authors, contributors, and community! 🌀🚀

We published an arXiv preprint entitled "OptunaHub: A Platform for Black-Box Optimization." The paper describes the motivation and the ecosystem overview. arxiv.org/abs/2510.02798…

Kaito Baba (@kAIto47802) added constraint handling for multi-objective optimization by GPSampler, a Gaussian process-based Bayesian optimization. It is especially beneficial when many expensive metrics need to be considered. Learn more in our article: medium.com/optuna/optuna-…

We recently published an arXiv article about a sample-efficient black-box combinatorial optimization for TPE by considering distance structures of categorical parameters. This feature is available via categorical_distance_func in TPESampler. arxiv.org/abs/2507.08053

Optuna v4.5 has been released! ⛏️GPSampler for constrained multi-objective optimization 🚀Significant speedup of TPESampler and plot_hypervolume_history 🦾CmaEsSampler now supports 1D search space 🐍The optunahub library is published on conda-forge github.com/optuna/optuna/…

Optuna v4.4 is now available as of June 16! This release introduces the Optuna MCP Server, our first LLM-based toolchain. We're already hard at work on Optuna v5, following our roadmap. For full details, check out our blog post. medium.com/optuna/announc…

See the release blog for more details. medium.com/optuna/announc…

Check out the release note! github.com/optuna/optuna/…

Optuna v4.4 has been released with various new features, bug fixes, and enhancements. 🚀Optuna MCP server, which is our first LLM intensive toolchain ✅Gaussian process-based algorithm now supports multi-objective optimization 🌀A lot of new features in OptunaHub

See the repository for more details of Optuna MCP server github.com/optuna/optuna-…

Optuna v5 will be built with the power of the community! We'd love your input and participation — check out the feedback form on the roadmap: forms.gle/wVwLCQ9g6st6AX…

See our blog for more details! medium.com/optuna/optuna-…

We've published the development roadmap for Optuna v5, the next major release! Scheduled for release next summer, v5 will focus on making Optuna even more powerful and user-friendly — especially at the intersection of generative AI and black-box optimization.