Statements (24)
| Predicate | Object |
|---|---|
| gptkbp:instanceOf |
gptkb:mathematical_optimization
|
| gptkbp:advantage |
handles non-stationary objectives
may require tuning of hyperparameters reduces learning rate for frequent parameters |
| gptkbp:application |
neural network training
|
| gptkbp:category |
gradient descent optimization
|
| gptkbp:feature |
adaptive learning rate
element-wise learning rate adjustment |
| gptkbp:implementedIn |
available in Keras
available in PyTorch available in TensorFlow |
| gptkbp:parameter |
learning rate
decay rate epsilon |
| gptkbp:proposedBy |
gptkb:Geoffrey_Hinton
2012 |
| gptkbp:relatedTo |
gptkb:Adam
gptkb:AdaGrad |
| gptkbp:updateRule |
divides learning rate by moving average of squared gradients
|
| gptkbp:usedIn |
gptkb:machine_learning
deep learning |
| gptkbp:bfsParent |
gptkb:Adam_optimizer
|
| gptkbp:bfsLayer |
5
|
| https://www.w3.org/2000/01/rdf-schema#label |
RMSProp
|