rfann 0.1.0

A pure Rust implementation of the Fast Artificial Neural Network (FANN) library
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
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265

use crate::{ActivationFunction, Connection, Layer, Network, NetworkBuilder, Neuron};
use approx::assert_relative_eq;

#[test]
fn test_connection_creation() {
    let conn = Connection::new(0, 1, 0.5_f32);
    assert_eq!(conn.from_neuron, 0);
    assert_eq!(conn.to_neuron, 1);
    assert_relative_eq!(conn.weight, 0.5_f32);
}

#[test]
fn test_connection_generic_f64() {
    let conn = Connection::new(0, 1, 0.5_f64);
    assert_eq!(conn.from_neuron, 0);
    assert_eq!(conn.to_neuron, 1);
    assert_relative_eq!(conn.weight, 0.5_f64);
}

#[test]
fn test_neuron_creation() {
    let neuron = Neuron::<f32>::new(ActivationFunction::Sigmoid, 1.0);
    assert_eq!(neuron.activation_function, ActivationFunction::Sigmoid);
    assert_relative_eq!(neuron.activation_steepness, 1.0);
    assert_relative_eq!(neuron.sum, 0.0);
    assert_relative_eq!(neuron.value, 0.0);
    assert!(neuron.connections.is_empty());
}

#[test]
fn test_neuron_add_connection() {
    let mut neuron = Neuron::<f32>::new(ActivationFunction::Sigmoid, 1.0);
    neuron.add_connection(0, 0.5);

    assert_eq!(neuron.connections.len(), 1);
    assert_eq!(neuron.connections[0].from_neuron, 0);
    assert_relative_eq!(neuron.connections[0].weight, 0.5);
}

#[test]
fn test_layer_creation() {
    let layer = Layer::<f32>::new(3, ActivationFunction::Sigmoid, 1.0);
    assert_eq!(layer.neurons.len(), 3);

    for neuron in &layer.neurons {
        assert_eq!(neuron.activation_function, ActivationFunction::Sigmoid);
        assert_relative_eq!(neuron.activation_steepness, 1.0);
    }
}

#[test]
fn test_layer_with_bias() {
    let layer = Layer::<f32>::with_bias(3, ActivationFunction::Sigmoid, 1.0);
    assert_eq!(layer.neurons.len(), 4); // 3 regular + 1 bias
    assert!(layer.has_bias());

    // Bias neuron should have value 1.0
    let bias_neuron = &layer.neurons[3];
    assert_relative_eq!(bias_neuron.value, 1.0);
}

#[test]
fn test_network_builder_basic() {
    let network: Network<f32> = NetworkBuilder::new()
        .input_layer(2)
        .hidden_layer(3)
        .output_layer(1)
        .build();

    assert_eq!(network.num_layers(), 3);
    assert_eq!(network.num_inputs(), 2);
    assert_eq!(network.num_outputs(), 1);

    // Check layer sizes (including bias neurons in input and hidden layers)
    assert_eq!(network.layers[0].neurons.len(), 3); // 2 input + 1 bias
    assert_eq!(network.layers[1].neurons.len(), 4); // 3 hidden + 1 bias
    assert_eq!(network.layers[2].neurons.len(), 1); // 1 output (no bias)
}

#[test]
fn test_network_builder_with_activation_functions() {
    let network: Network<f32> = NetworkBuilder::new()
        .input_layer(2)
        .hidden_layer_with_activation(3, ActivationFunction::ReLU, 1.0)
        .output_layer_with_activation(1, ActivationFunction::Linear, 1.0)
        .build();

    // Check activation functions
    let hidden_layer = &network.layers[1];
    for neuron in &hidden_layer.neurons[..3] {
        // Exclude bias neuron
        assert_eq!(neuron.activation_function, ActivationFunction::ReLU);
    }

    let output_layer = &network.layers[2];
    assert_eq!(
        output_layer.neurons[0].activation_function,
        ActivationFunction::Linear
    );
}

#[test]
fn test_network_fully_connected() {
    let network: Network<f32> = NetworkBuilder::new()
        .input_layer(2)
        .hidden_layer(3)
        .output_layer(1)
        .connection_rate(1.0)
        .build();

    // Check connections
    // Input layer (2 neurons + 1 bias) should connect to all 3 hidden neurons
    let hidden_layer = &network.layers[1];
    for i in 0..3 {
        // For each hidden neuron (excluding bias)
        let neuron = &hidden_layer.neurons[i];
        assert_eq!(neuron.connections.len(), 3); // Connected to 2 input + 1 bias
    }

    // Hidden layer (3 neurons + 1 bias) should connect to output neuron
    let output_layer = &network.layers[2];
    let output_neuron = &output_layer.neurons[0];
    assert_eq!(output_neuron.connections.len(), 4); // Connected to 3 hidden + 1 bias
}

#[test]
fn test_network_sparse_connections() {
    let network: Network<f32> = NetworkBuilder::new()
        .input_layer(10)
        .hidden_layer(10)
        .output_layer(10)
        .connection_rate(0.5)
        .build();

    // With 50% connection rate, we should have roughly half the connections
    let total_connections = network.total_connections();
    let max_connections = 11 * 10 + 11 * 10; // (10+1) * 10 + (10+1) * 10

    assert!(total_connections > (max_connections as f32 * 0.3) as usize); // Allow some variance
    assert!(total_connections < (max_connections as f32 * 0.7) as usize);
}

#[test]
fn test_network_multiple_hidden_layers() {
    let network: Network<f32> = NetworkBuilder::new()
        .input_layer(2)
        .hidden_layer(4)
        .hidden_layer(3)
        .hidden_layer(2)
        .output_layer(1)
        .build();

    assert_eq!(network.num_layers(), 5);
    assert_eq!(network.layers[1].neurons.len(), 5); // 4 + 1 bias
    assert_eq!(network.layers[2].neurons.len(), 4); // 3 + 1 bias
    assert_eq!(network.layers[3].neurons.len(), 3); // 2 + 1 bias
}

#[test]
fn test_network_generic_f64() {
    let network: Network<f64> = NetworkBuilder::new()
        .input_layer(2)
        .hidden_layer(3)
        .output_layer(1)
        .build();

    assert_eq!(network.num_layers(), 3);
    assert_eq!(network.num_inputs(), 2);
    assert_eq!(network.num_outputs(), 1);
}

#[test]
fn test_network_run_forward_pass() {
    let mut network: Network<f32> = NetworkBuilder::new()
        .input_layer(2)
        .hidden_layer(2)
        .output_layer(1)
        .build();

    // Set some weights manually for testing
    // This is a simple test, in real usage weights would be trained
    let hidden_layer = &mut network.layers[1];
    for neuron in &mut hidden_layer.neurons[..2] {
        neuron.connections.clear();
        neuron.add_connection(0, 0.5); // From input 0
        neuron.add_connection(1, 0.5); // From input 1
        neuron.add_connection(2, 0.0); // From bias
    }

    let output_layer = &mut network.layers[2];
    output_layer.neurons[0].connections.clear();
    output_layer.neurons[0].add_connection(3, 1.0); // From hidden 0
    output_layer.neurons[0].add_connection(4, 1.0); // From hidden 1
    output_layer.neurons[0].add_connection(5, 0.0); // From bias

    let input = vec![1.0, 1.0];
    let output = network.run(&input);

    assert_eq!(output.len(), 1);
    // The actual value depends on activation function implementation
}

#[test]
fn test_activation_functions() {
    // Test that all activation function variants are available
    let _sigmoid = ActivationFunction::Sigmoid;
    let _tanh = ActivationFunction::Tanh;
    let _relu = ActivationFunction::ReLU;
    let _linear = ActivationFunction::Linear;
    let _gaussian = ActivationFunction::Gaussian;
    let _sigmoid_symmetric = ActivationFunction::SigmoidSymmetric;
    let _elliot = ActivationFunction::Elliot;
    let _elliot_symmetric = ActivationFunction::ElliotSymmetric;
    let _sin = ActivationFunction::Sin;
    let _cos = ActivationFunction::Cos;
    let _sin_symmetric = ActivationFunction::SinSymmetric;
    let _cos_symmetric = ActivationFunction::CosSymmetric;
    let _threshold = ActivationFunction::Threshold;
    let _threshold_symmetric = ActivationFunction::ThresholdSymmetric;
    let _linear_piece = ActivationFunction::LinearPiece;
    let _linear_piece_symmetric = ActivationFunction::LinearPieceSymmetric;
    let _relu_leaky = ActivationFunction::ReLULeaky;
}

#[test]
fn test_network_get_weights() {
    let network: Network<f32> = NetworkBuilder::new()
        .input_layer(2)
        .hidden_layer(2)
        .output_layer(1)
        .build();

    let weights = network.get_weights();
    let expected_weights = 3 * 2 + 3; // (2+1 bias) * 2 hidden + (2+1 bias) * 1 output
    assert_eq!(weights.len(), expected_weights);
}

#[test]
fn test_network_set_weights() {
    let mut network: Network<f32> = NetworkBuilder::new()
        .input_layer(2)
        .hidden_layer(2)
        .output_layer(1)
        .build();

    let new_weights = vec![0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9];
    assert!(network.set_weights(&new_weights).is_ok());

    let retrieved_weights = network.get_weights();
    for (i, &weight) in retrieved_weights.iter().enumerate() {
        assert_relative_eq!(weight, new_weights[i]);
    }
}

#[test]
fn test_network_set_weights_wrong_size() {
    let mut network: Network<f32> = NetworkBuilder::new()
        .input_layer(2)
        .hidden_layer(2)
        .output_layer(1)
        .build();

    let wrong_weights = vec![0.1, 0.2]; // Too few weights
    assert!(network.set_weights(&wrong_weights).is_err());
}