irithyll-core 1.0.0

Core types, training engine, and inference for irithyll streaming ML — no_std + alloc, histogram binning, Hoeffding trees, SGBT ensembles, drift detection, f32 + int16 packed formats
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46

//! Spiking Neural Networks with online e-prop learning.
//!
//! This module provides LIF (Leaky Integrate-and-Fire) neurons with integer
//! arithmetic (Q1.14 fixed-point), delta spike encoding, and the e-prop
//! three-factor learning rule for online training.
//!
//! # Architecture
//!
//! The SNN processes streaming data through three stages:
//!
//! 1. **Delta encoding** -- raw features are converted to spike trains via
//!    temporal differencing (UP/DOWN spike pairs per feature)
//! 2. **Recurrent spiking layer** -- LIF neurons with e-prop eligibility
//!    traces and feedback alignment for weight updates
//! 3. **Readout** -- non-spiking leaky integrators whose membrane potentials
//!    serve as continuous output predictions
//!
//! # Fixed-Point Arithmetic
//!
//! All neuron computations use Q1.14 fixed-point (i16 where 16384 = 1.0).
//! Multiply-accumulate operations use i32 intermediates to prevent overflow.
//! This enables deployment on integer-only hardware (Cortex-M0+, RISC-V).
//!
//! # Components
//!
//! - [`lif`] -- LIF neuron step and surrogate gradient functions
//! - [`spike_encoding`] -- Delta spike encoder for continuous-to-spike conversion
//! - [`eprop`] -- e-prop three-factor learning rule functions
//! - [`readout`] -- Non-spiking leaky integrator readout neuron
//! - [`network_fixed`] -- Complete SNN combining all components

pub mod astrocyte;
pub mod eprop;
pub mod lif;
pub mod network_fixed;
pub mod readout;
pub mod spike_encoding;

pub use eprop::{
    compute_learning_signal_fixed, update_eligibility_fixed, update_output_weights_fixed,
    update_pre_trace_fixed, update_weights_fixed,
};
pub use lif::{lif_step, surrogate_gradient_pwl, Q14_HALF, Q14_ONE};
pub use network_fixed::{Precision, SpikeNetFixed};
pub use readout::ReadoutNeuron;
pub use spike_encoding::DeltaEncoderFixed;