Struct GrowingNeuralGas 

pub struct GrowingNeuralGas {
    pub lr_winner: f64,
    pub lr_neighbor: f64,
    pub max_age: usize,
    pub insert_interval: usize,
    pub alpha: f64,
    pub beta: f64,
    pub max_units: usize,
    pub max_steps: usize,
    pub seed: u64,
}

Growing Neural Gas — adaptive topology competitive learning.

The GNG starts with two prototype units and grows by inserting new nodes between the highest-error nodes. Edges age and are removed when their age exceeds max_age. No a priori unit count is needed.

Reference: Fritzke (1995).

Fields

lr_winner: f64

Learning rate for the winner unit. Default 0.1.

lr_neighbor: f64

Learning rate for topological neighbours. Default 0.01.

max_age: usize

Maximum edge age before removal. Default 50.

insert_interval: usize

Steps between node insertions. Default 100.

alpha: f64

Error reduction factor applied to nodes after insertion. Default 0.5.

beta: f64

Global error decay per step. Default 0.995.

max_units: usize

Upper bound on the number of units. Default 200.

max_steps: usize

Total training steps. Default 5000.

seed: u64

RNG seed. Default 42.

Implementations

impl GrowingNeuralGas

pub fn new( lr_winner: f64, lr_neighbor: f64, max_age: usize, insert_interval: usize, alpha: f64, beta: f64, max_units: usize, max_steps: usize, seed: u64, ) -> Self

Create a new GNG with all hyperparameters specified.

pub fn fit(&self, data: ArrayView2<'_, f64>) -> Result<GrowingNeuralGasModel>

Fit the GNG model to data.

Arguments

• data – Data matrix (n_samples, n_features).

Returns

A GrowingNeuralGasModel with the learned topology and assignments.

Trait Implementations

impl Clone for GrowingNeuralGas

fn clone(&self) -> GrowingNeuralGas

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for GrowingNeuralGas

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

Formats the value using the given formatter.

impl Default for GrowingNeuralGas

fn default() -> Self

Returns the “default value” for a type.