Statements (48)
Predicate | Object |
---|---|
gptkbp:instanceOf |
Optimization algorithm
|
gptkbp:advantage |
Efficient for expensive evaluations
Requires fewer function evaluations Works with noisy functions |
gptkbp:alternativeTo |
Evolutionary algorithms
Gradient-based optimization Grid search Random search |
gptkbp:field |
gptkb:Mathematics
Statistics Machine learning |
gptkbp:firstDescribed |
1970s
Jonas Mockus |
https://www.w3.org/2000/01/rdf-schema#label |
Bayesian Optimization
|
gptkbp:limitation |
Computationally expensive for large datasets
Model selection can be challenging Scales poorly with high dimensions |
gptkbp:notableFor |
gptkb:robot
A/B testing Engineering design Drug discovery Automated machine learning (AutoML) Neural network hyperparameter tuning |
gptkbp:relatedTo |
Active learning
Multi-armed bandit problem Expected improvement Global optimization Probability of improvement Sequential model-based optimization Surrogate model Upper confidence bound |
gptkbp:software |
gptkb:Spearmint
gptkb:SMAC gptkb:Optuna gptkb:Hyperopt GPyOpt scikit-optimize BayesianOptimization (Python package) |
gptkbp:usedFor |
Black-box optimization
Expensive function optimization Hyperparameter tuning |
gptkbp:uses |
gptkb:Gaussian_process
Probabilistic model Random forest Tree-structured Parzen Estimator Acquisition function |
gptkbp:bfsParent |
gptkb:Bayesian_Learning
|
gptkbp:bfsLayer |
7
|