BoFire for BoTorch 0.9.4

What's Changed

• Fix number of experiments condition for zero experiments by @jduerholt in #266
• Smaller model dumps by @jduerholt in #267
• Added outlier detection tutorial by @swagataroy123 in #262
• Add generic benchmark module by @jduerholt in #269
• Add log single-objective ACQFs by @jduerholt in #271
• Bayesian optimization over molecules by @swagataroy123 in #268
• Change random seed behavior by @jduerholt in #276
• Sync the output of surrogate.predict and strategy.predict by @jduerholt in #282
• Fix infeasible cost calculation for categorical inputs by @jduerholt in #281
• Refactor get acquisition in SOBO strategies and adapt to new way of handling constraints by @gmancino in #275
• Feature/experiment validation by @jduerholt in #289
• Implement possibility to run hyperparameter opts in the strategy by @jduerholt in #287
• Refactor the candidates/experiments validators. by @jduerholt in #291
• Add possibility to compute feature importance over the lengthscales in a SingleTaskGPSurrogate by @jduerholt in #293
• Fix for categorical features that are fixed in fully categorical BO by @simonsung06 in #295
• Mapping happens outside of runner by @bertiqwerty in #296
• Add polynomial kernel by @dlinzner-bcs in #298
• 286 add surrogate models that can extrapolate by @dlinzner-bcs in #299
• Iterative branch and bound by @dlinzner-bcs in #302
• Updated rdkit fragment/descriptor list by @simonsung06 in #306
• Allow for max_active=#(variables in constraint) in n_choose_k_constraints_as_bounds() by @Osburg in #304
• Add output_scaler to mlp by @simonsung06 in #305
• Add multi-objective log ACQFs by @jduerholt in #308

New Contributors

• @gmancino made their first contribution in #275

Full Changelog: v0.0.7...v0.0.8

BoFire for BoTorch 0.9.2

This release includes a BoFire version which requires BoTorch 0.9.2. In addition the following things are new:

What's Changed

• relaxing error to a warning by @ufukguenes in #257
• Outlier detection in predictive strategies by @swagataroy123 in #235
• added calibration metric by @swagataroy123 in #253
• CategoricalMolecularFeature by @jduerholt in #260
• Fix for to and from categorical variable encoding by @simonsung06 in #261
• DoE for categorical features by @ufukguenes in #259
• stratified kfold for cv by @simonsung06 in #263
• Make compatible with latest botorch release and main branch by @jduerholt in #265

New Contributors

• @ufukguenes made their first contribution in #257

Full Changelog: v0.0.6...v0.0.7

BoFire for BoTorch 0.9.1

This release includes a BoFire version which is compatible with BoTorch 0.9.1 and therefore requires at least Python 3.9.

BoFire for Python 3.8

This release is the only BoFire release that supports Python 3.8. It uses BoTorch 0.8.5. Later releases will require at least Python 3.9 and use BoTorch >= 0.9.1.