NAG class

Implements the Nesterov Accelerated Gradient (NAG) optimizer.

NAG is an improvement over standard momentum. While standard momentum combines the previous velocity with the current gradient, NAG first makes a "lookahead" step in the direction of the velocity and then calculates the gradient from that future position. This correction prevents the optimizer from overshooting the minimum and often leads to faster convergence.

Analogy 🧠

Think of a ball rolling down a hill:

• Momentum is a heavy ball that builds up speed and rolls past small bumps.
• NAG is a smarter ball that looks ahead at the slope just before its next move. It can slow down if the ground is about to rise, preventing it from overshooting the bottom of the valley.

Example

var optimizer = NAG(model.parameters, learningRate: 0.01, momentum: 0.9);

Inheritance

• Object
• Optimizer
• NAG

Constructors

NAG(List<Tensor> parameters, {required double learningRate, double momentum = 0.9})

Properties

hashCode → int

The hash code for this object.

no setter, inherited

learningRate → double

The step size for the gradient updates.

final, inherited

momentum → double

final

parameters → List<Tensor>

The list of model parameters (weights and biases) that this optimizer will update.

final, inherited

runtimeType → Type

A representation of the runtime type of the object.

no setter, inherited

Methods

noSuchMethod(Invocation invocation) → dynamic

Invoked when a nonexistent method or property is accessed.

inherited

step() → void

Performs a single optimization step using an efficient NAG update rule.

override

toString() → String

A string representation of this object.

inherited

zeroGrad() → void

Resets the gradients of all parameters to zero.

inherited

Operators

operator ==(Object other) → bool

The equality operator.

inherited