Struct LinearAttackModel 

pub struct LinearAttackModel {
    pub weights: Vec<Vec<f64>>,
    pub bias: Vec<f64>,
}

A simple linear model f(x) = W·x + b used primarily for testing attacks.

Fields

weights: Vec<Vec<f64>>

Weight matrix: weights[i] is the i-th output row (length = n_inputs).

bias: Vec<f64>

Bias vector (length = n_outputs).

Implementations

impl LinearAttackModel

pub fn new( weights: Vec<Vec<f64>>, bias: Vec<f64>, ) -> Result<Self, AdversarialError>

Construct a new linear model, validating that all rows have the same length.

Trait Implementations

impl AttackModel for LinearAttackModel

fn input_gradient(&self, _input: &[f64], output_grad: &[f64]) -> Vec<f64>

Exact analytical gradient for a linear model: ∂(g·Wx)/∂x = Wᵀ·g.

fn forward(&self, input: &[f64]) -> Vec<f64>

Forward pass: given an input slice, return predictions (logits or probs).