| Entrance | What is it used for? | Starting Point |
| Machine Learning | Fit a predictive model for a quantitative or qualitative Y based on several X variables. | An existing data grid. |
| Optimization | Find the setting that works best several answers at a time, based on the fitted models. | One or more models. |
| Bayesian Optimization | Achieve optimal results with a minimum number of tests: the module suggests initial tests and subsequent ones as needed. | Nothing—factors and their limits. |
| Mix Optimization | The same principle applies to a formulation whose components total 100 %. | Nothing—just the components and their terminals. |
💡 The first two entries are working based on collected data, the last two lead the acquisition. A useful tip: If you can still decide which tests to run, sequential optimization modules will save you from having to run those tests; if the tests have already been run, that’s machine learning. It is possible to incorporate previously conducted experiments into Bayesian optimization, provided that these experiments adequately cover the domain. Similarly, it is possible to start with existing mixtures, provided that these mixtures satisfy the condition «Sum of proportions = 1.»
⚠️ All four entries require a project selected. Machine Learning and Optimization also read a grid ; the two sequential optimization modules are standalone and store their data within the study itself.