Struct DeepPolyNeuron 

pub struct DeepPolyNeuron {
    pub lb_coeffs: Vec<f64>,
    pub lb_bias: f64,
    pub ub_coeffs: Vec<f64>,
    pub ub_bias: f64,
    pub concrete_lb: f64,
    pub concrete_ub: f64,
}

A single-neuron abstract element in the DeepPoly domain.

Each neuron y_i has a concrete value bounded by a linear expression: lb_coeffs · x + lb_bias ≤ y_i ≤ ub_coeffs · x + ub_bias where x are the (input) neurons from a preceding layer.

Fields

lb_coeffs: Vec<f64>

Coefficients of the lower linear bound (over preceding layer neurons)

lb_bias: f64

Constant of the lower linear bound

ub_coeffs: Vec<f64>

Coefficients of the upper linear bound

ub_bias: f64

Constant of the upper linear bound

concrete_lb: f64

Concrete lower bound (after back-substitution)

concrete_ub: f64

Concrete upper bound (after back-substitution)

Implementations

impl DeepPolyNeuron

pub fn constant(val: f64, input_dim: usize) -> Self

Create a constant neuron (no uncertainty).

pub fn abstract_relu(&self) -> Self

Apply abstract ReLU: case analysis on the concrete bounds.

pub fn contains_concrete(&self, val: f64) -> bool

Check if a concrete value is within the abstract bounds.

Trait Implementations

impl Clone for DeepPolyNeuron

fn clone(&self) -> DeepPolyNeuron

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for DeepPolyNeuron

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

Formats the value using the given formatter.