vlachosgroup / nextorch Goto Github PK

Nice work on NEXTorch! Seems like a compelling idea, and some difficulty with implementing human-in-the-loop optimization via Ax is what led me here.

Something very important for my use-case is inequality constraints for composition-based Bayesian optimization (also prevalence-based, fractional-based, etc., i.e. everything needs to sum to 1). For example:

A + B + C <= 1

where A, B, and C are parameters to be optimized.

For me, there would technically also be D, the final parameter is determined automatically outside of the optimization loop via:

D = 1 - sum([A, B, C])

In other words, safe to ignore D, but relevant for the context of a composition-based optimization.

Is it possible to specify an inequality constraint within the current implementation of NEXTorch?

I tried to create a conda environment where Ax and nextorch could play together, but I am not so lucky.
I installed ax-platform first.

nextorch 0.4.0 requires botorch<=0.4.0, but you have botorch 0.9.5 which is incompatible.
nextorch 0.4.0 requires gpytorch<=1.4, but you have gpytorch 1.11 which is incompatible.

Please update nextorch to make that possible.

Once I want to run your example in the supporting information of your paper, I cannot import PFR_yield. I also cannot install this package.

Thank you for a great tool and this is probably something that is non-urgent as the core code still seems to work.

When trying to produce heat map and surface plots using the template I end up with a uniform value across all values:

I initially thought that there was a problem with my code, but when running one of your notebooks (10_PFR_mixed_type_inputs.ipynb) I see the exact same behaviour:

My assumption is that this is probably just a dependency issue. The optimal values still seem to be correct and so it doesn't seem to be a problem that affects the underlying data processing.