The development of iterative design of experiments based on a Bayesian approach is gaining interest as shown in these two articles:
- Shields, B. J., Stevens, J., Li, J., Parasram, M., Damani, F., Alvarado, J. I. M., ... & Doyle, A. G. (2021). Bayesian reaction optimization as a tool for chemical synthesis. Nature, 590(7844), 89-96.
- Greenhill, S., Rana, S., Gupta, S., Vellanki, P., & Venkatesh, S. (2020). Bayesian optimization for adaptive experimental design: a review. IEEE access, 8, 13937-13948.
A JMP add-in is available to implement this approach (https://community.jmp.com/t5/JMP-Add-Ins/Bayesian-optimization-add-in/ta-p/496785). Unfortunately, the add-in lacks some important features such as:
- the use of several model types (in particular the bootstrap forest)
- the management of discrete variables
- the management of experimental constraints
- the possibility to run several experiments per iteration
It would be very appreciable and useful to have such a platform available, especially in the fields of exploratory research for the design of new materials.
@Florent_M thank you for the suggesting. We will investigate this further and get back to you when have more information to share.
Hi @Florent_M, thank you for your suggestion! We have captured your request and will take it under consideration.
Hello Florent_M . The new Bayesian Optimization platform in JMP 19 (under Specialized Modeling), should satisfy all of your requests in this wish (except possibly the use of different model types). The platform uses a GaSP (Gaussian Stochastic Process) model approach. Please test out the new platform in JMP 19 and let us know what you think. Thanks!
