Struct PositionalEncoding 

pub struct PositionalEncoding {
    pub num_levels: usize,
}

Positional encoding: maps a scalar or vector into a higher-dimensional Fourier feature space.

For a scalar p, the encoding is: [sin(2^0 pi p), cos(2^0 pi p), ..., sin(2^{L-1} pi p), cos(2^{L-1} pi p)]

For an N-dimensional vector the encoding is applied independently to each component, producing a 2 * L * N-dimensional output.

Fields

num_levels: usize

Number of frequency levels.

Implementations

impl PositionalEncoding

pub fn new(num_levels: usize) -> Self

Create a new positional encoding with num_levels frequency bands.

pub fn output_dim(&self, input_dim: usize) -> usize

Output dimensionality for an input of input_dim components.

pub fn encode_scalar(&self, value: f64) -> Vec<f64>

Encode a single scalar value.

pub fn encode(&self, values: &[f64]) -> Vec<f64>

Encode a vector of values, concatenating per-component encodings.

Trait Implementations

impl Clone for PositionalEncoding

fn clone(&self) -> PositionalEncoding

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for PositionalEncoding

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

Formats the value using the given formatter.