search

modAL-python / modAL

A modular active learning framework for Python
https://modAL-python.github.io/
MIT License
2.2k stars 324 forks source link

[FEATURE REQUEST] Can we use Regressor in Ranked batch-mode sampling? #118

Open rzc331 opened 3 years ago

rzc331 commented 3 years ago

In the documentation, only classifier is supported for Ranked batch-mode sampling. I am wondering whether regressor is supported? And if yes, how should we do it? Thanks in advance!

cosmic-cortex commented 3 years ago

Hi! As far as I know, ranked batch-mode learning officially doesn't support regression, but after taking a quick look at how it is done, it should work with regression as well. Although this was not tested in the original paper (https://www.sciencedirect.com/science/article/abs/pii/S0020025516313949).

Have you tried plugging in a Gaussian Process regressor?

rzc331 commented 3 years ago

Thank you so much for replying!

I am actually a beginner on active learning. I tried naively plugging in a GP regressor and it does not work. I also searched some other packages and did not find one supporting this feature, other than GPyOpt which is out of maintainence unfrotunately.

I really like modAL because of its beginner friendliness, which makes things a lot easier compared with other packages of the same kind. It'd be best if modAL could be implemented with Batch BO regression feature, which is highly useful in experimental science related field.

Thanks again!