Struct DeltaEncoderFixed 

pub struct DeltaEncoderFixed { /* private fields */ }

Available on crate feature alloc only.

Delta spike encoder for fixed-point inputs.

Converts continuous-valued features (Q1.14 i16) into binary spike trains by detecting temporal changes exceeding per-feature thresholds.

Layout

For n_features input features, the output has 2 * n_features channels:

• Channel 2*i = positive spike (feature increased)
• Channel 2*i + 1 = negative spike (feature decreased)

Implementations

impl DeltaEncoderFixed

pub fn new(n_features: usize, threshold: i16) -> Self

Create a new delta encoder with a uniform threshold for all features.

Arguments

• n_features – number of raw input features
• threshold – uniform spike threshold in Q1.14

pub fn new_per_feature(thresholds: Vec<i16>) -> Self

Create a new delta encoder with per-feature thresholds.

Arguments

• thresholds – one threshold per input feature, in Q1.14

pub fn n_features(&self) -> usize

Number of raw input features.

pub fn n_output_channels(&self) -> usize

Number of output spike channels (always 2 * n_features).

pub fn encode(&mut self, input: &[i16], out_spikes: &mut [u8])

Encode an input vector into a spike buffer.

The out_spikes buffer must have length >= 2 * n_features. It will be filled with 0s and 1s: even indices for positive spikes, odd for negative.

On the very first call, no spikes are emitted (the encoder needs a previous value to compute deltas). The previous values are stored for the next call.

Panics

Panics if input.len() != n_features or out_spikes.len() < 2 * n_features.

pub fn reset(&mut self)

Reset the encoder to its initial state.

The next call to encode will produce no spikes (warmup step).