Struct cerebral::selforganizing::SelforganizingNetwork

pub struct SelforganizingNetwork<A, T, R, L>where
    A: Adaptable<NeuralLayer, R>,
    T: Topological<NeuralLayer>,
    R: Responsive<NeuralLayer>,
    L: Trainable<NeuralLayer, A, R>,{
    pub neurons: NeuralLayer,
    pub adaptivity: A,
    pub topology: T,
    pub responsiveness: R,
    pub training: L,
}

Struct that implements structural composition

Fields

neurons: NeuralLayer

needs to be nested to share it with the algorithms

adaptivity: A

Algorithm for adaptivity

topology: T

Algorithm related to topology

responsiveness: R

Algorithm to feature pattern matching and lateral inhibition

training: L

Algorithm related to batch processing

Trait Implementations

impl<A, T, R, B> Neural for SelforganizingNetwork<A, T, R, B>where A: Adaptable<NeuralLayer, R>, T: Topological<NeuralLayer>, R: Responsive<NeuralLayer>, B: Trainable<NeuralLayer, A, R>,

fn get_lateral(&self) -> &Array2<f64>

fn get_lateral_mut(&mut self) -> &mut Array2<f64>

fn set_lateral(&mut self, lateral: Array2<f64>)

fn get_patterns(&self) -> &Array2<f64>

fn get_patterns_mut(&mut self) -> &mut Array2<f64>

fn set_patterns(&mut self, patterns: Array2<f64>)

impl<A, T, R, B> Selforganizing for SelforganizingNetwork<A, T, R, B>where A: Adaptable<NeuralLayer, R>, T: Topological<NeuralLayer>, R: Responsive<NeuralLayer>, B: Trainable<NeuralLayer, A, R>,

fn init_lateral(&mut self)

Init the lateral connections according to network type

fn get_lateral_distance(&mut self, index: usize) -> Array2<f64>

Get the distance/connection between a selected neuron and the rest of the layer

fn get_best_matching(&mut self, pattern: &ArrayView1<'_, f64>) -> usize

Get the best matching neuron given a pattern

fn adapt(&mut self, pattern: &ArrayView1<'_, f64>, influence: f64, rate: f64)

Adapt the layer to an input pattern. Note this consumes the current later and returns a new created (zero-copy)

fn train(&mut self, patterns: &ArrayView2<'_, f64>)