PAPER: Matter (?) article for using CLSLabs-Light as a teaching and prototyping demo

[sgbaird]

Adaptive experimentation for the physical sciences: A more efficient alternative to design of experiments

Tutorials and templates for teaching advanced optimization and autonomous experimentation in the physical sciences

CLSLabs-Light: A low-cost self-driving laboratory for teaching autonomous experimentation topics in the physical sciences

CLSLabs-Light as a teaching and prototyping platform for advanced optimization and autonomous experimentation in the physical sciences

Artifacts:

• Figure: Annotated CLSLabs:Light hardware and summary figures
• Table: summary of tutorial notebooks
• Table: summary of the basic, fidelity, batch, high-dimensional, (Olympus) category, and scalability results
• Figure: a comparison between grid search, random search, and Bayesian optimization for experimental and simulated results
  • subfigures showing the search behavior as the optimization progresses
  • subfigure showing the solution space (mean and standard deviation)
• Figure: Pareto front exploration behavior of qEHVI vs. scalarized objectives
• Figure: efficiency comparison between single- and multi-fidelity (continuous) optimization
• Figure: efficiency comparison between sequential vs. batch vs. asynchronous optimization
• Figure: efficiency comparison between single- and multi-fidelity (discrete) optimization
• Figure: efficiency comparison between high-dimensional Bayesian optimization and traditional Bayesian optimization
• Figure: comparisons between Olympus benchmarking algorithms
• Figure: flowchart for setting up and storing in a database, and later retrieval
• Figure: flowchart for the use of experiment planning software
• Figure: Runtime and efficiency comparison for scalable Bayesian optimization vs. traditional Bayesian optimization
• Figure: Cloud control - MQTT communication between device and client
• Figure: teaching figure for how the simulations are set up