Discussions
Solve problems, and share tips and tricks with other JMP users.
Accepted Solutions
Hi @aatw,
I just tested to create a small DoE with only categorical factors and a random response, everything seems to be accepted in the BO platform, no error messages or warning:
When entering in the platform, an automatic recommendation is done, like for continuous input variables:
So no problem to use categorical factors only with BO JMP Platform; in other situations/packages, you may need to use a "Bandit Optimization" method, which is more adapted to choosing levels from a discrete set of candidates. More infos on the Ax webpage about Bandit Optimization vs. Bayesian Optimization: https://ax.dev/docs/0.5.0/banditopt/
Note that I only answer regarding the feasibility of doing it in JMP, not about any success rate or adapted strategy regarding your use case.
Best,
Hi @aatw,
I just tested to create a small DoE with only categorical factors and a random response, everything seems to be accepted in the BO platform, no error messages or warning:
When entering in the platform, an automatic recommendation is done, like for continuous input variables:
So no problem to use categorical factors only with BO JMP Platform; in other situations/packages, you may need to use a "Bandit Optimization" method, which is more adapted to choosing levels from a discrete set of candidates. More infos on the Ax webpage about Bandit Optimization vs. Bayesian Optimization: https://ax.dev/docs/0.5.0/banditopt/
Note that I only answer regarding the feasibility of doing it in JMP, not about any success rate or adapted strategy regarding your use case.
Best,