neuburn 0.1.0

Spiking neural network framework on top of Burn — Rust-native alternative to snntorch
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
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76

//! Recurrent Synaptic: LIF with synaptic conductance traces and recurrent weights.
//! I = W_in @ x + W_rec @ spike_prev; syn' = α·syn + I; mem' = β·mem + syn.

use super::spike_straight_through;
use crate::state::{apply_reset, NeuronState, ResetMode};
use crate::surrogate::Surrogate;
use burn::module::{Ignored, Module, Param};
use burn::nn::Linear;
use burn::nn::LinearConfig;
use burn::tensor::activation::sigmoid;
use burn::tensor::backend::Backend;
use burn::tensor::Tensor;

#[derive(Module, Debug)]
pub struct RSynaptic<B: Backend> {
    pub linear_in: Linear<B>,
    pub linear_rec: Linear<B>,
    pub beta: Param<Tensor<B, 1>>,
    pub alpha_syn: Param<Tensor<B, 1>>,
    pub threshold: f32,
    pub reset_mode: Ignored<ResetMode>,
}

impl<B: Backend> RSynaptic<B> {
    pub fn new(
        device: &B::Device,
        in_features: usize,
        out_features: usize,
        beta_init: f32,
        alpha_syn_init: f32,
        threshold: f32,
        reset_mode: ResetMode,
    ) -> Self {
        let linear_in = LinearConfig::new(in_features, out_features).init(device);
        let linear_rec = LinearConfig::new(out_features, out_features).init(device);
        let beta_raw = (beta_init / (1.0 - beta_init.max(0.01f32))).ln();
        let beta = Param::from_tensor(Tensor::from_floats([beta_raw], device));
        let alpha_raw = (alpha_syn_init / (1.0 - alpha_syn_init.max(0.01f32))).ln();
        let alpha_syn = Param::from_tensor(Tensor::from_floats([alpha_raw], device));
        Self {
            linear_in,
            linear_rec,
            beta,
            alpha_syn,
            threshold,
            reset_mode: Ignored(reset_mode),
        }
    }

    pub fn step<S: Surrogate>(
        &self,
        input: Tensor<B, 2>,
        spike_prev: Tensor<B, 2>,
        state: &NeuronState<B>,
        surrogate: &S,
    ) -> (Tensor<B, 2>, NeuronState<B>) {
        let i_in = self.linear_in.forward(input);
        let i_rec = self.linear_rec.forward(spike_prev);
        let current = i_in + i_rec;
        let syn = state.syn.as_ref().unwrap_or(&state.mem);
        let b = sigmoid(self.beta.val().clone()).reshape([1, 1]).expand(state.mem.dims());
        let a = sigmoid(self.alpha_syn.val().clone())
            .reshape([1, 1])
            .expand(state.mem.dims());
        let syn_new = syn.clone().mul(a) + current;
        let mem_new = state.mem.clone().mul(b) + syn_new.clone();
        let spike = spike_straight_through(mem_new.clone(), self.threshold, surrogate);
        let spike_hard = spike.clone().greater_equal_elem(0.5).float();
        let mem_reset = apply_reset(mem_new, &spike_hard, self.threshold, self.reset_mode.0);
        let new_state = NeuronState {
            mem: mem_reset,
            syn: Some(syn_new),
        };
        (spike, new_state)
    }
}