Nesterov accelerated gradient
GPTKB entity
Statements (23)
| Predicate | Object |
|---|---|
| gptkbp:instanceOf |
mathematical optimization
|
| gptkbp:alsoKnownAs |
Nesterov momentum
|
| gptkbp:application |
training neural networks
convex optimization |
| gptkbp:category |
first-order optimization method
|
| gptkbp:convergesTo |
O(1/k^2)
|
| gptkbp:feature |
faster convergence for smooth convex functions
lookahead gradient uses momentum term |
| gptkbp:form |
x_{k+1} = y_k - abla f(y_k)
y_{k+1} = x_{k+1} + eta_k (x_{k+1} - x_k) |
| https://www.w3.org/2000/01/rdf-schema#label |
Nesterov accelerated gradient
|
| gptkbp:improves |
gradient descent
momentum method |
| gptkbp:proposedBy |
Yurii Nesterov
|
| gptkbp:relatedTo |
gptkb:Adam_optimizer
gptkb:RMSprop momentum |
| gptkbp:usedIn |
gptkb:machine_learning
deep learning |
| gptkbp:yearProposed |
1983
|
| gptkbp:bfsParent |
gptkb:Momentum_optimizer
|
| gptkbp:bfsLayer |
6
|