Struct DeepKernelBuilder 

pub struct DeepKernelBuilder { /* private fields */ }

Fluent builder for Deep Kernel networks.

The builder records layer widths and activations in order. The first width is the input dimension (set via Self::input_dim). Every subsequent hidden layer appends a new width and its activation. The terminating Self::output_dim records the final width and the output-layer activation.

Implementations

impl DeepKernelBuilder

pub fn new() -> Self

Empty builder — no layers configured yet.

pub fn input_dim(self, dim: usize) -> Self

Set the input dimension. Must be called exactly once before the first hidden or output layer.

pub fn hidden_layer(self, width: usize, activation: Activation) -> Self

Append a hidden layer with the given width and activation. Ignored if the builder has already been closed with Self::output_dim — the builder reports that as an error at build time via KernelError::InvalidParameter.

pub fn output_dim(self, width: usize, activation: Activation) -> Self

Finalise the topology by appending the output layer. Must be called exactly once.

pub fn seed(self, seed: u64) -> Self

Set the RNG seed used for Xavier initialisation.

pub fn build_extractor(&self) -> Result<MLPFeatureExtractor>

Produce an owned MLPFeatureExtractor for the configured topology, without a base kernel. Useful when the caller wants to combine the MLP with a kernel that needs additional plumbing.

pub fn build<K: Kernel>( self, base: K, ) -> Result<DeepKernel<MLPFeatureExtractor, K>>

Finalise the builder against a base kernel and produce a fully wired DeepKernel.

Trait Implementations

impl Clone for DeepKernelBuilder

fn clone(&self) -> DeepKernelBuilder

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for DeepKernelBuilder

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

Formats the value using the given formatter.

impl Default for DeepKernelBuilder

fn default() -> DeepKernelBuilder

Returns the “default value” for a type.