Struct ParallelSpikingNeuron 

pub struct ParallelSpikingNeuron {
    pub weights: Vec<f64>,
    pub history: Vec<f64>,
    pub v_threshold: f64,
    pub hidden: f64,
}

k-order sliding PSN — Fang et al. (2023), NeurIPS.

Streaming form of the PSN family (paper Eqs. 14–15):

H[t] = sum_{i=0}^{k-1} W_i * X[t-k+1+i],  X[j] = 0 for j < 0
S[t] = Theta(H[t] - v_threshold)

weights[k-1] multiplies the newest input; the sum accumulates sequentially from i = 0 so every backend reproduces the same binary64 result bit-for-bit. Theta(0) = 1 per the paper. No PSN variant has a reset: firing never clears the input history. The paper trains W and v_threshold; the uniform 1/k defaults are repository defaults.

Fields

weights: Vec<f64>

history: Vec<f64>

v_threshold: f64

hidden: f64

Implementations

impl ParallelSpikingNeuron

pub fn new(kernel_size: usize, v_threshold: f64) -> Self

Construct with the uniform repository default weights 1/k.

fn valid(&self) -> bool

pub fn try_step(&mut self, current: f64) -> Result<i32, &'static str>

Advance one step after validating the input and configuration.

Computes the hidden state on a candidate window and commits only on success: a non-finite input, an invalid configuration, or a non-finite hidden state returns Err with the pre-step state preserved exactly.

pub fn step(&mut self, current: f64) -> i32

Fail-closed wrapper for legacy callers: returns 0 on any rejected input without mutating state.

pub fn reset(&mut self)

Clear the retained inputs, preserving weights and threshold.

Trait Implementations

impl Clone for ParallelSpikingNeuron

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ParallelSpikingNeuron

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

Formats the value using the given formatter.