Skip to content

Latest commit

 

History

History
33 lines (27 loc) · 1.37 KB

README.md

File metadata and controls

33 lines (27 loc) · 1.37 KB

PrefInt

The PrefInt package provides a way of integrating data via preference learning. For more details, please check our online manuscript: https://arxiv.org/abs/1910.11516

Update

We have provided a new repository DPDI that integrating data via neural network with higher efficiency. Please refer to: PrefIntNN

Python version: 3.7
!The code was recently rewrote from python 2.7 to python 3.7.

Example

For generating pairwise preference via traindata and add preference with new point:
This is for adding single new datapoint during iteration in optimization. For adding more datapoints in one step, please extend the traindata to regenerate the pairs.

from PrefInt.preference_generate import generate_pair
GN = generate_pair(traindata)
prefs = GN.firstgen()

newprefs = GN.addnew(newdata)

For training the pairwise preference:

from PrefInt.ibo.gaussianprocess import PrefGaussianProcess
from PrefInt.ibo.gaussianprocess.kernel import GaussianKernel_iso

kernel = GaussianKernel_iso(np.array([18.0]))
GP = PrefGaussianProcess(kernel)
GP.addPreferences(prefs)

For more examples, please check examples and utility.

License

The PrefInt package is licensed under the MIT "Expat" License