dial-0.1.0

DIAL (Distributed INTERSECT Active Learning): An INTERSECT service that provides Bayesian optimization and active learning

Dial v0.1

Initial public release of Dial. Dial is an active learning tool for the INTERSECT Ecosystem implemented using the INTERSECT SDK. The current features allow simple active learning workflows using Gaussian process surrogates

Initial features:

• Bayesian optimization using backends from scikit-learn and GPax
• Support for expected improvement, variance, and upper confidence bound acquisition functions
• Support for random and latin hypercube selection of initial samples
• Example use cases for automated and manual integration into a scientific workflow
• Preprocessing options (scale-by-mean and logorithmic transformations)
• Integration into INTERSECT using the INTERSECT SDK

Full Changelog: https://github.com/INTERSECT-DIAL/dial/commits/v0.1