Home

Summit is a package for optimization of chemical processes. It works by suggesting new experiments based on past experiments or randomized initial designs. There are five main parts:

• Domains: Specifications of the optimization problem. For example, for reaction optimization, you might have an objective of maximizing yield by changing the stoichiometry of two reactants and reaction temperature.

• Strategies: The algorithm used for suggesting new experiments based on past experiments. TSEMO is the main strategy included currently.

• Datasets: Experimental or physical data that can be used in the optimization. For example, if you have descriptors of solvents, these could be included as a dataset.

• Models: Statistical or analytical models that act as "surrogates" of the thing being optimized. For example, you might use a linear model to represent the relationship between stoichiometry and yield.

The tutorial is the best way to learn how to use summit.