Leave-Cluster-Out Cross-Validation Is Appropriate for Scoring Functions Derived from Diverse Protein Data Sets.
Christian KramerPeter GedeckPublished in: J. Chem. Inf. Model. (2010)
Keyphrases
- cross validation
- scoring functions
- data sets
- training set
- model selection
- unseen data
- hyperparameters
- scoring function
- support vector
- generalization error
- error estimates
- clustering algorithm
- cross validated
- classification accuracy
- nearest neighbor classifiers
- high dimensional data
- nearest neighbor
- reinforcement learning
- feature extraction
- feature selection
- machine learning