Struct GepSymRegression 

pub struct GepSymRegression<F: FunctionSet> { /* private fields */ }

A symbolic-regression BatchFitnessFn for GEP populations.

Holds the target dataset (input rows and expected outputs) and the alphabet to decode against. evaluate_batch pulls the population to host and, in parallel across rows (rayon), decodes each chromosome to an ExpressionTree and scores it by mean squared error. Decoding is deterministic, so the row-parallel order does not affect results.

Implementations

impl<F: FunctionSet> GepSymRegression<F>

pub fn new( alphabet: Alphabet<F>, genome_len: usize, inputs: Vec<Vec<f32>>, targets: Vec<f32>, ) -> Self

Builds the fitness function from an alphabet, the genome length, the input rows, and the matching expected outputs.

Panics

Panics if:

• inputs is empty — an empty dataset would make MSE 0.0 / 0.0 = NaN fitness (n_points == 0), silently poisoning the population.
• inputs and targets differ in length.
• any input row does not have exactly alphabet.n_vars entries — a short row is otherwise silently zero-padded by ExpressionTree::eval (inputs.get(i).unwrap_or(0.0)), so the regression would quietly fit the wrong target.

Trait Implementations

impl<B: Backend, F: FunctionSet> BatchFitnessFn<B, Tensor<B, 2, Int>> for GepSymRegression<F>

fn evaluate_batch( &mut self, population: &Tensor<B, 2, Int>, device: &<B as BackendTypes>::Device, ) -> Tensor<B, 1>

Evaluates every member of population and returns a fitness tensor in the objective’s natural value space (no hand-negation).

fn sense(&self) -> ObjectiveSense

The optimisation direction of this objective.

impl<F: Clone + FunctionSet> Clone for GepSymRegression<F>

fn clone(&self) -> GepSymRegression<F>

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<F: Debug + FunctionSet> Debug for GepSymRegression<F>

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

Formats the value using the given formatter.