human-augmented Bayesian optimized spectral recommendation system: A "human-in-the-loop Automated Experiment framework"

MLExchange Project funded by Berkeley Lab, award number 107514

Brief Problem Description

• Here we have a image data, where X is the input location of the image

• Each location in the image, we have a spectral data, from where user select if the data is good/bad. We define target spectral from user votes and feature preference on good sampled spectral

• The goal is to build a optimization (BO) model where we adaptively learn the target features/properties of the material sample and simultaneously maximize the structural sim of spectral to the target at the current state of BO process.

Different architectures of BO spectral recommendation system:

BOSRS with standard GP and 2D co-ordinate as input X (in torch) https://github.com/arpanbiswas52/varTBO/blob/main/BO_spectral_(Notebookversion).ipynb

BOSRS with standard GP and high-dim image patch as input X (in torch) https://github.com/arpanbiswas52/varTBO/blob/main/BO(image_patch)_(Notebookversion).ipynb

dKLBOSRS with dKLGP (deep learning kernel GP) and high-dim image patch as input X (in numpy) https://github.com/arpanbiswas52/varTBO/blob/main/dKLBO_spectral_(Notebookversion).ipynb https://github.com/arpanbiswas52/varTBO/blob/main/dKLBO_spectral_MicroscopyVersionv2.ipynb

Note* the deep learning kernel (dKL) function in this architecture is developed by Maxim Ziatdinov (in AtomAI Python library). Details on dkL model can be found here https://atomai.readthedocs.io/en/latest/atomai_models.html#atomai.models.dklGPR

Please feel free to use these notebooks for your research, however, please cite the github repository

Please let me know if you have questions, find any bugs, issues, or want another option/feature added at either biswasar@ornl.gov, arpanbiswas52@gmail.com.