Struct EncoderNetwork 

pub struct EncoderNetwork<T: Float + Debug + Send + Sync + 'static> { /* private fields */ }

Encoder network for feature extraction

Implementations

impl<T: Float + Debug + Send + Sync + 'static> EncoderNetwork<T>

pub fn forward(&self, features: &Array1<T>) -> Result<Array1<T>>

Embed a raw feature vector.

Each layer computes Wᵗx + b (the weight matrix is stored (input_dim, output_dim)), and the activation is applied to every layer except the last — a prototypical-network embedding is the final linear map’s output, not a rectified version of it.

A feature vector that does not match the first layer’s input width is zero-padded or truncated to fit, which is documented rather than silent: the caller’s raw task features have no fixed width, and refusing them outright would make the encoder unusable on real task data.

Errors

Returns Err when the network has no layers, or when two consecutive layers disagree on width.

pub fn embedding_width(&self) -> Result<usize>

Width of the embedding this network produces.

Errors

Returns Err when the network has no layers.

Trait Implementations

impl<T: Debug + Float + Debug + Send + Sync + 'static> Debug for EncoderNetwork<T>

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

Formats the value using the given formatter.