Trait SoftmaxKernelV1 

pub trait SoftmaxKernelV1 {
    // Required method
    fn softmax(&self, x: &[f32]) -> Vec<f32>;
}

Contract trait for softmax-kernel-v1 v1.0.0.

Softmax kernel — numerically stable exponential normalization Reference: Bridle (1990) Training Stochastic Model Recognition Algorithms as Networks Reference: Milakov & Gimelshein (2018) Online normalizer calculation for softmax

Implementors must provide all 1 equation(s). Missing method = compile error. Wrong signature = compile error.

Required Methods

fn softmax(&self, x: &[f32]) -> Vec<f32>

softmax: σ(x)_i = exp(x_i - max(x)) / Σ_j exp(x_j - max(x)) Domain: x ∈ ℝ^n, n ≥ 1 Codomain: σ(x) ∈ (0,1)^n Invariant: Σ σ(x)_i = 1.0 (normalization) Invariant: σ(x)_i > 0 for all i (strict positivity) Invariant: argmax(σ(x)) = argmax(x) (order preservation)

Implementors