Enum GcStrategy 

pub enum GcStrategy {
    LayerWise,
    Global,
    PerRow,
    PerColumn,
}

Gradient centralization strategy.

Variants

LayerWise

Centralize each layer’s gradients independently (most common). For each parameter matrix: g = g - mean(g)

Global

Centralize all gradients globally (experimental). Compute global mean across all parameters: g_all = g_all - mean(g_all)

PerRow

Centralize per row (for weight matrices). For weight matrix: g[i,:] = g[i,:] - mean(g[i,:])

PerColumn

Centralize per column (for weight matrices). For weight matrix: g[:,j] = g[:,j] - mean(g[:,j])

Trait Implementations

impl Clone for GcStrategy

fn clone(&self) -> GcStrategy

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for GcStrategy

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

Formats the value using the given formatter.

impl Default for GcStrategy

fn default() -> GcStrategy

Returns the “default value” for a type.

impl PartialEq for GcStrategy

fn eq(&self, other: &GcStrategy) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Copy for GcStrategy

impl StructuralPartialEq for GcStrategy