Struct RMSProp 

pub struct RMSProp<B>
where
    B: Backend,
{ /* private fields */ }

RMSProp optimizer.

Adapts the learning rate per-parameter using a running average of squared gradients.

Default hyperparameters

• lr: 0.01
• alpha: 0.99 (smoothing constant)
• epsilon: 1e-8
• momentum: 0.0
• weight_decay: 0.0

Implementations

impl<B> RMSProp<B>

where B: Backend,

pub fn new(params: Vec<Tensor<B>>, lr: f64) -> RMSProp<B>

Create a new RMSProp optimizer with default hyperparameters.

pub fn alpha(self, alpha: f64) -> RMSProp<B>

Set the smoothing constant α (default: 0.99).

pub fn epsilon(self, eps: f64) -> RMSProp<B>

Set ε (numerical stability, default: 1e-8).

pub fn momentum(self, momentum: f64) -> RMSProp<B>

Set momentum factor (default: 0).

pub fn weight_decay(self, wd: f64) -> RMSProp<B>

Set weight decay / L2 penalty (default: 0).

pub fn params(&self) -> &[Tensor<B>]

Access current parameters.

pub fn params_mut(&mut self) -> &mut Vec<Tensor<B>>

Mutable access to current parameters.

Trait Implementations

impl<B> Optimizer<B> for RMSProp<B>

where B: Backend,

fn step(&mut self, grads: &GradStore<B>) -> Result<Vec<Tensor<B>>, Error>

Perform one optimization step.

fn learning_rate(&self) -> f64

Return the current learning rate.

fn set_learning_rate(&mut self, lr: f64)

Set a new learning rate (for learning rate scheduling).

impl<B> Stateful for RMSProp<B>

where B: Backend,

fn state_dict(&self) -> OptimizerState

Export the optimizer’s internal state as a serializable dictionary.

fn load_state_dict(&mut self, state: &OptimizerState) -> Result<(), Error>

Restore the optimizer’s internal state from a previously saved dictionary.