ncps-rust 0.1.1

Neural Circuit Policies - Sparse RNNs inspired by C. elegans, implemented in Rust with Burn
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
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167

//! Integration tests for the LTC Cell

use burn::backend::NdArray;
use burn::tensor::{Distribution, Tensor};
use ncps::cells::LTCCell;
use ncps::wirings::{AutoNCP, FullyConnected};

type Backend = NdArray<f32>;

fn create_test_cell() -> LTCCell<Backend> {
    let device = Default::default();
    let wiring = FullyConnected::new(10, Some(5), 1234, true);

    LTCCell::new(&wiring, Some(8), &device)
        .with_ode_unfolds(6)
        .with_epsilon(1e-8)
}

#[test]
fn test_ltc_cell_creation() {
    let device = Default::default();
    let wiring = FullyConnected::new(10, Some(5), 1234, true);
    let cell = LTCCell::<Backend>::new(&wiring, Some(8), &device);

    assert_eq!(cell.state_size(), 10);
    assert_eq!(cell.motor_size(), 5);
    assert_eq!(cell.sensory_size(), 8);
}

#[test]
fn test_ltc_cell_forward() {
    let device = Default::default();
    let cell = create_test_cell();

    let batch_size = 4;
    let inputs = Tensor::<Backend, 2>::zeros([batch_size, 8], &device);
    let states = Tensor::<Backend, 2>::zeros([batch_size, 10], &device);
    let elapsed_time = Tensor::<Backend, 1>::ones([batch_size], &device);

    let (output, new_state) = cell.forward(inputs, states, elapsed_time);

    // Output is motor_size, state is state_size
    assert_eq!(output.dims(), [batch_size, 5]);
    assert_eq!(new_state.dims(), [batch_size, 10]);
}

#[test]
fn test_ltc_state_change() {
    let device = Default::default();
    let cell = create_test_cell();

    let inputs = Tensor::<Backend, 2>::random([2, 8], Distribution::Uniform(-1.0, 1.0), &device);
    let states = Tensor::<Backend, 2>::zeros([2, 10], &device);
    let elapsed_time = Tensor::<Backend, 1>::full([2], 1.0, &device);

    let (output, new_state) = cell.forward(inputs.clone(), states.clone(), elapsed_time.clone());

    // State should have changed
    let state_diff = new_state.abs().mean().into_scalar();
    assert!(state_diff > 0.0, "State should change after forward pass");
}

#[test]
fn test_ltc_synapse_count() {
    let device = Default::default();
    let wiring = FullyConnected::new(10, Some(5), 1234, true);
    let cell = LTCCell::<Backend>::new(&wiring, Some(8), &device);

    // Synapse counts should be as expected
    assert_eq!(cell.synapse_count(), 100); // state_size^2
    assert_eq!(cell.sensory_synapse_count(), 80); // sensory_size * state_size
}

#[test]
fn test_ltc_parameter_shapes() {
    let device = Default::default();
    let wiring = FullyConnected::new(5, Some(3), 1234, true);
    let cell = LTCCell::<Backend>::new(&wiring, Some(4), &device);

    // Check dimensions
    assert_eq!(cell.state_size(), 5);
    assert_eq!(cell.motor_size(), 3);
    assert_eq!(cell.sensory_size(), 4);
}

#[test]
fn test_ltc_different_time_steps() {
    let device = Default::default();
    let cell = create_test_cell();

    let inputs = Tensor::<Backend, 2>::ones([2, 8], &device);
    let states = Tensor::<Backend, 2>::zeros([2, 10], &device);

    // Small time step
    let ts_small = Tensor::<Backend, 1>::full([2], 0.1, &device);
    let (out1, state1) = cell.forward(inputs.clone(), states.clone(), ts_small);

    // Large time step
    let ts_large = Tensor::<Backend, 1>::full([2], 1.0, &device);
    let (out2, state2) = cell.forward(inputs, states, ts_large);

    // Different time steps should produce different results
    let diff = (state1 - state2).abs().mean().into_scalar();
    assert!(
        diff > 0.01,
        "Different time steps should produce different states"
    );
}

#[test]
fn test_ltc_with_ncp_wiring() {
    use ncps::wirings::AutoNCP;

    let device = Default::default();
    let wiring = AutoNCP::new(32, 8, 0.5, 22222);
    let cell = LTCCell::<Backend>::new(&wiring, Some(16), &device);

    assert_eq!(cell.state_size(), 32);
    assert_eq!(cell.motor_size(), 8);

    let inputs = Tensor::<Backend, 2>::zeros([2, 16], &device);
    let states = Tensor::<Backend, 2>::zeros([2, 32], &device);
    let elapsed_time = Tensor::<Backend, 1>::ones([2], &device);

    let (output, new_state) = cell.forward(inputs, states, elapsed_time);

    assert_eq!(output.dims(), [2, 8]);
    assert_eq!(new_state.dims(), [2, 32]);
}

#[test]
fn test_ltc_different_batch_sizes() {
    let device = Default::default();
    let wiring = FullyConnected::new(10, Some(5), 1234, true);
    let cell = LTCCell::<Backend>::new(&wiring, Some(8), &device);

    for batch_size in [1, 4, 16, 32] {
        let inputs = Tensor::<Backend, 2>::zeros([batch_size, 8], &device);
        let states = Tensor::<Backend, 2>::zeros([batch_size, 10], &device);
        let elapsed_time = Tensor::<Backend, 1>::ones([batch_size], &device);

        let (output, new_state) = cell.forward(inputs, states, elapsed_time);

        assert_eq!(output.dims(), [batch_size, 5]);
        assert_eq!(new_state.dims(), [batch_size, 10]);
    }
}

#[test]
fn test_ltc_with_random_inputs() {
    let device = Default::default();
    let wiring = FullyConnected::new(10, Some(5), 1234, true);
    let cell = LTCCell::<Backend>::new(&wiring, Some(8), &device);

    let inputs = Tensor::<Backend, 2>::random([4, 8], Distribution::Uniform(-2.0, 2.0), &device);
    let states = Tensor::<Backend, 2>::zeros([4, 10], &device);
    let elapsed_time = Tensor::<Backend, 1>::ones([4], &device);

    let (output, new_state) = cell.forward(inputs, states, elapsed_time);

    assert_eq!(output.dims(), [4, 5]);
    assert_eq!(new_state.dims(), [4, 10]);

    // Output should be finite
    let output_sum = output.sum().into_scalar();
    assert!(output_sum.is_finite(), "Output should be finite");
}