Struct TrainableBox 

pub struct TrainableBox {
    pub mu: Vec<f32>,
    pub delta: Vec<f32>,
    pub dim: usize,
}

A trainable box embedding with learnable parameters.

Uses reparameterization to ensure min <= max:

• min = mu - exp(delta)/2
• max = mu + exp(delta)/2

This ensures boxes are always valid (min <= max).

Fields

mu: Vec<f32>

Mean position in each dimension (d-dimensional vector).

delta: Vec<f32>

Log-width in each dimension (width = exp(delta)).

dim: usize

Dimension

Implementations

impl TrainableBox

pub fn new(mu: Vec<f32>, delta: Vec<f32>) -> Self

Create a new trainable box.

Arguments

• mu - Mean position (center of box)
• delta - Log-width (width = exp(delta))

The box will have:

• min = mu - exp(delta) / 2
• max = mu + exp(delta) / 2

pub fn from_vector(vector: &[f32], init_width: f32) -> Self

Initialize from a vector embedding.

Creates a small box around the vector with initial width init_width.

pub fn to_box(&self) -> BoxEmbedding

Convert to a BoxEmbedding (for inference).

impl TrainableBox

pub fn update_amsgrad( &mut self, grad_mu: &[f32], grad_delta: &[f32], state: &mut AMSGradState, )

Update box parameters using AMSGrad optimizer.

Trait Implementations

impl Clone for TrainableBox

fn clone(&self) -> TrainableBox

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for TrainableBox

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for TrainableBox

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for TrainableBox

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.