Struct EntropyRegularization 

pub struct EntropyRegularization<A: Float + FromPrimitive + Debug> {
    pub lambda: A,
    pub epsilon: A,
    pub reg_type: EntropyRegularizerType,
}

Entropy regularization for probability distributions

This regularizer can either encourage high entropy (more uniform distributions) or low entropy (more peaked distributions) depending on the selected regularizer type.

It’s commonly used in reinforcement learning algorithms, semi-supervised learning, and some classification tasks where controlling the certainty of outputs is desired.

Fields

lambda: A

Regularization strength

epsilon: A

Small value for numerical stability

reg_type: EntropyRegularizerType

Type of entropy regularization

Implementations

impl<A: Float + FromPrimitive + Debug + Send + Sync> EntropyRegularization<A>

pub fn new(lambda: A, regtype: EntropyRegularizerType) -> Self

Create a new entropy regularization

Arguments

• lambda - Regularization strength coefficient
• reg_type - Type of entropy regularization (maximize or minimize)

Returns

An entropy regularization with default epsilon

pub fn new_with_epsilon( lambda: A, epsilon: A, regtype: EntropyRegularizerType, ) -> Self

Create a new entropy regularization with custom epsilon

Arguments

• lambda - Regularization strength coefficient
• epsilon - Small value for numerical stability
• reg_type - Type of entropy regularization (maximize or minimize)

Returns

An entropy regularization with custom epsilon

pub fn calculate_entropy<S, D>(&self, probs: &ArrayBase<S, D>) -> A

where S: Data<Elem = A>, D: Dimension,

Calculate the entropy of a probability distribution

Arguments

• probs - Probability distribution (should sum to 1 along the appropriate axis)

Returns

The entropy value

Trait Implementations

impl<A: Clone + Float + FromPrimitive + Debug> Clone for EntropyRegularization<A>

fn clone(&self) -> EntropyRegularization<A>

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<A: Debug + Float + FromPrimitive + Debug> Debug for EntropyRegularization<A>

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

Formats the value using the given formatter.

impl<A, D> Regularizer<A, D> for EntropyRegularization<A>

where A: Float + ScalarOperand + Debug + FromPrimitive + Send + Sync, D: Dimension,

fn apply(&self, params: &Array<A, D>, gradients: &mut Array<A, D>) -> Result<A>

Apply regularization to parameters and gradients

fn penalty(&self, params: &Array<A, D>) -> Result<A>

Compute the regularization penalty value

impl<A: Copy + Float + FromPrimitive + Debug> Copy for EntropyRegularization<A>