Olympus

Overview

Olympus is a research software framework for comparing optimization algorithms and experiment-planning methods. Its role is to provide benchmark problems and a consistent interface for planners, rather than to operate laboratory hardware directly. The README identifies chemistry-oriented applications involving experimental constraints, robust process optimization and transfer learning for reaction optimization.

Benchmark workflows can use domains containing continuous, discrete and categorical parameters. The documented collection includes 23 experiment-planning algorithms, 33 experimentally derived benchmarks and 33 analytical test functions. These give users several ways to assess planners against experimental data or mathematical objectives. The framework also accepts custom optimization algorithms and custom datasets; the latter can support training models for additional benchmark problems. Plotting and analysis facilities help users inspect benchmark experiments and communicate their results.

Inputs therefore include parameter-domain definitions, selected benchmark problems and, for extensions, user-provided datasets or algorithms. The documented outputs include benchmark experiment results and associated visualizations, although the supplied excerpt does not specify result schemas, dataset formats or individual planner APIs. Those details need to be checked in the linked documentation before implementing a workflow.

The README lists Python, numpy and pandas as baseline dependencies, with additional libraries needed for particular modules and objects. Its stated Python minimum is an installation requirement, not evidence of currently tested compatibility. The linked sources describe available capabilities but provide no comparative performance results or evidence that benchmark findings transfer directly to a particular laboratory process.

Key Features

• Defines optimization domains using continuous, discrete and categorical parameters.
• Exposes 23 experiment-planning algorithms through a consistent interface, according to the project README.
• Provides 33 experimentally derived benchmarks and 33 analytical optimization test functions.
• Supports integration of custom optimization algorithms for benchmarking.
• Supports custom datasets that can be used to train models for additional benchmarks.
• Offers plotting and analysis options for visualizing benchmark experiments.

Use Cases

• Suggested evaluation: compare experiment planners on experimentally derived benchmarks before selecting candidates for a chemistry optimization study.
• Suggested evaluation: assess planner behavior on domains containing continuous, discrete and categorical variables relevant to an experimental design.
• Suggested evaluation: integrate a new optimization algorithm and compare it with existing planners on analytical test functions.
• Suggested evaluation: use a project-specific dataset to train a custom benchmark model and investigate planner behavior before considering laboratory deployment.

How to Use

1. Read the Olympus documentation to identify the domain, benchmark and planner interfaces needed for your study. Check the detailed APIs rather than assuming input or output formats from the overview.
2. Consult the official README for installation options. It lists pip, conda and source installation, plus Python, numpy and pandas as baseline dependencies; individual modules may require additional libraries.
3. Explore the getting-started Colab notebook to understand an example workflow before adapting it to your own benchmark question.
4. Select an analytical or experimentally derived benchmark and define the relevant parameter types. For a custom dataset or planner, follow the extension guidance in the documentation and confirm the required representations.
5. Run your intended comparison and inspect results with the documented plotting and analysis facilities. Record the chosen benchmark, planner and dependencies; treat conclusions about a specific laboratory process as a separate evaluation question.