Statements (26)
| Predicate | Object |
|---|---|
| gptkbp:instanceOf |
regularization technique
|
| gptkbp:alternativeTo |
early stopping
L1 regularization L2 regularization |
| gptkbp:appliesTo |
training phase
|
| gptkbp:category |
machine learning regularization
|
| gptkbp:citation |
gptkb:JMLR_2014_paper_"Dropout:_A_Simple_Way_to_Prevent_Neural_Networks_from_Overfitting"
|
| gptkbp:commonDropoutRate |
0.5
|
| gptkbp:effect |
improves generalization
reduces interdependent learning among neurons |
| https://www.w3.org/2000/01/rdf-schema#label |
Dropout (neural networks)
|
| gptkbp:implementedIn |
gptkb:TensorFlow
gptkb:Keras gptkb:PyTorch |
| gptkbp:method |
randomly sets units to zero during training
|
| gptkbp:notAppliedDuring |
inference phase
|
| gptkbp:parameter |
dropout rate
|
| gptkbp:proposedBy |
gptkb:Geoffrey_Hinton
2014 |
| gptkbp:purpose |
prevent overfitting
|
| gptkbp:relatedTo |
bagging
ensemble learning |
| gptkbp:usedIn |
gptkb:artificial_neural_networks
deep learning |
| gptkbp:bfsParent |
gptkb:Geoffrey_E._Hinton
|
| gptkbp:bfsLayer |
6
|