Mlrintermbo: Model-Based Optimization for 'Mlr3' Through 'MlrMBO'

MCML Authors

Martin Binder

→ Group Bernd Bischl
 Statistical Learning and Data Science

Abstract

The 'mlrMBO' package can ordinarily not be used for optimization within 'mlr3', because of incompatibilities of their respective class systems. 'mlrintermbo' offers a compatibility interface that provides 'mlrMBO' as an 'mlr3tuning' 'Tuner' object, for tuning of machine learning algorithms within 'mlr3', as well as a 'bbotk' 'Optimizer' object for optimization of general objective functions using the 'bbotk' black box optimization framework. The control parameters of 'mlrMBO' are faithfully reproduced as a 'paradox' 'ParamSet'.

manual

---

SW Package

Jan. 2021

---

Authors

M. Binder

---

Links

URL GitHub

---

Research Area

 A1 | Statistical Foundations & Explainability