Struct AcceleratedGradientSVM 

pub struct AcceleratedGradientSVM {
    pub momentum: f64,
    pub learning_rate: f64,
    pub method: AcceleratedMethod,
    /* private fields */
}

Accelerated Gradient SVM Solver

Implements accelerated gradient descent methods for SVM optimization, including Nesterov’s accelerated gradient method and FISTA. These methods achieve faster convergence rates than standard gradient descent.

Nesterov’s method uses momentum to accelerate convergence: y_{k+1} = x_k - γ∇f(x_k) x_{k+1} = y_{k+1} + β(y_{k+1} - y_k)

Reference: Nesterov, Y. (2013). Introductory lectures on convex optimization.

Fields

momentum: f64

Momentum parameter (typically 0.9)

learning_rate: f64

Learning rate schedule

method: AcceleratedMethod

Accelerated method type

Implementations

impl AcceleratedGradientSVM

pub fn new(config: AdvancedOptimizationConfig) -> Self

Create a new accelerated gradient SVM solver

pub fn with_momentum(self, momentum: f64) -> Self

Set the momentum parameter

pub fn with_learning_rate(self, learning_rate: f64) -> Self

Set the learning rate

pub fn with_method(self, method: AcceleratedMethod) -> Self

Set the accelerated method type

pub fn fit( &mut self, x: &Array2<f64>, y: &Array1<f64>, ) -> Result<OptimizationResult>

Fit the SVM using accelerated gradient optimization

pub fn predict( &self, x: &Array2<f64>, result: &OptimizationResult, ) -> Result<Array1<f64>>

Make predictions

pub fn decision_function( &self, x: &Array2<f64>, result: &OptimizationResult, ) -> Result<Array1<f64>>

Compute decision function

Trait Implementations

impl Clone for AcceleratedGradientSVM

fn clone(&self) -> AcceleratedGradientSVM

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for AcceleratedGradientSVM

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.