rustyml 0.11.0

A high-performance machine learning & deep learning library in pure Rust, offering ML algorithms and neural network support
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

#![cfg(feature = "neural_network")]

use approx::assert_abs_diff_eq;
use ndarray::Array;
use rustyml::neural_network::Tensor;
use rustyml::neural_network::layer::TrainingParameters;
use rustyml::neural_network::layer::activation_layer::linear::Linear;
use rustyml::neural_network::layer::activation_layer::relu::ReLU;
use rustyml::neural_network::layer::activation_layer::sigmoid::Sigmoid;
use rustyml::neural_network::layer::activation_layer::tanh::Tanh;
use rustyml::neural_network::layer::dense::Dense;
use rustyml::neural_network::loss_function::mean_squared_error::MeanSquaredError;
use rustyml::neural_network::neural_network_trait::Layer;
use rustyml::neural_network::optimizer::sgd::SGD;
use rustyml::neural_network::sequential::Sequential;

#[test]
fn test_dense_forward_pass_dimensions() {
    // Test forward propagation dimension correctness
    let mut dense = Dense::new(4, 3, ReLU::new()).unwrap();
    let input = Array::ones((2, 4)).into_dyn(); // batch_size=2, input_dim=4

    let output = dense.forward(&input).unwrap();
    assert_eq!(output.shape(), &[2, 3]); // batch_size=2, output_dim=3
    println!(
        "Forward pass dimension test passed: {:?} -> {:?}",
        input.shape(),
        output.shape()
    );
}

#[test]
fn test_dense_activation_functions() {
    // Test effects of different activation functions
    let input = Array::from_shape_vec((1, 2), vec![1.0, -1.0])
        .unwrap()
        .into_dyn();

    // Test ReLU activation function
    let mut dense_relu = Dense::new(2, 2, ReLU::new()).unwrap();
    let output_relu = dense_relu.forward(&input);

    // Test Sigmoid activation function
    let mut dense_sigmoid = Dense::new(2, 2, Sigmoid::new()).unwrap();
    let output_sigmoid = dense_sigmoid.forward(&input);

    // Test Tanh activation function
    let mut dense_tanh = Dense::new(2, 2, Tanh::new()).unwrap();
    let output_tanh = dense_tanh.forward(&input).unwrap();

    println!("ReLU output: {:?}", output_relu);
    println!("Sigmoid output: {:?}", output_sigmoid);
    println!("Tanh output: {:?}", output_tanh);

    // Verify Sigmoid output is in (0,1) range
    let sigmoid_output = output_sigmoid.unwrap();
    let sigmoid_2d = sigmoid_output
        .as_standard_layout()
        .into_dimensionality::<ndarray::Ix2>()
        .unwrap();
    for value in sigmoid_2d.iter() {
        assert!(
            *value > 0.0 && *value < 1.0,
            "Sigmoid output should be in (0,1) range: {}",
            value
        );
    }

    // Verify Tanh output is in (-1,1) range
    let tanh_2d = output_tanh
        .as_standard_layout()
        .into_dimensionality::<ndarray::Ix2>()
        .unwrap();
    for value in tanh_2d.iter() {
        assert!(
            *value > -1.0 && *value < 1.0,
            "Tanh output should be in (-1,1) range: {}",
            value
        );
    }
}

#[test]
fn test_dense_learning_capability() {
    // Test Dense layer learning capability - simple linear regression problem
    // y = 2*x1 + 3*x2 + 1

    // Create training data
    let x_train = Array::from_shape_vec(
        (4, 2),
        vec![
            1.0, 2.0, // y = 2*1 + 3*2 + 1 = 9
            2.0, 1.0, // y = 2*2 + 3*1 + 1 = 8
            3.0, 3.0, // y = 2*3 + 3*3 + 1 = 16
            0.0, 1.0, // y = 2*0 + 3*1 + 1 = 4
        ],
    )
    .unwrap()
    .into_dyn();

    let y_train = Array::from_shape_vec((4, 1), vec![9.0, 8.0, 16.0, 4.0])
        .unwrap()
        .into_dyn();

    // Build model
    let mut model = Sequential::new();
    model.add(Dense::new(2, 1, Linear::new()).unwrap());
    model.compile(SGD::new(0.01).unwrap(), MeanSquaredError::new());

    // Record initial prediction
    let initial_prediction = model.predict(&x_train).unwrap();
    let initial_loss = calculate_mse(&initial_prediction, &y_train);
    println!("Initial loss: {:.4}", initial_loss);

    // Train model
    model.fit(&x_train, &y_train, 1000).unwrap();

    // Test final prediction
    let final_prediction = model.predict(&x_train).unwrap();
    let final_loss = calculate_mse(&final_prediction, &y_train);
    println!("Final loss: {:.4}", final_loss);
    println!("Predictions: {:?}", final_prediction);
    println!("Ground truth: {:?}", y_train);

    // Verify model actually learned something (loss should be significantly reduced)
    assert!(
        final_loss < initial_loss * 0.1,
        "Model should learn patterns, loss reduced from {:.4} to {:.4}",
        initial_loss,
        final_loss
    );

    // Verify prediction accuracy
    let pred_2d = final_prediction
        .as_standard_layout()
        .into_dimensionality::<ndarray::Ix2>()
        .unwrap();
    let target_2d = y_train
        .as_standard_layout()
        .into_dimensionality::<ndarray::Ix2>()
        .unwrap();

    for i in 0..pred_2d.nrows() {
        let pred_val = pred_2d[[i, 0]];
        let true_val = target_2d[[i, 0]];
        assert_abs_diff_eq!(pred_val, true_val, epsilon = 1.0);
    }
}

#[test]
fn test_dense_batch_processing() {
    // Test batch processing capability
    let batch_sizes = vec![1, 5, 10, 32];

    for batch_size in batch_sizes {
        let input = Array::ones((batch_size, 4)).into_dyn();
        let mut dense = Dense::new(4, 3, ReLU::new()).unwrap();

        let output = dense.forward(&input).unwrap();
        assert_eq!(
            output.shape()[0],
            batch_size,
            "Batch size should remain unchanged"
        );
        assert_eq!(
            output.shape()[1],
            3,
            "Output feature dimension should be correct"
        );

        println!(
            "Batch size {} test passed: {:?} -> {:?}",
            batch_size,
            input.shape(),
            output.shape()
        );
    }
}

#[test]
fn test_dense_parameter_count() {
    // Test parameter count correctness
    let dense = Dense::new(10, 5, ReLU::new()).unwrap();
    let expected_params = 10 * 5 + 5; // weights + bias
    assert_eq!(
        dense.param_count(),
        TrainingParameters::Trainable(expected_params)
    );
    println!(
        "Parameter count test passed: {} parameters",
        expected_params
    );

    // Test different layer sizes
    let sizes = vec![(1, 1), (100, 50), (784, 128), (512, 10)];
    for (input_dim, output_dim) in sizes {
        let layer = Dense::new(input_dim, output_dim, Linear::new()).unwrap();
        let expected = input_dim * output_dim + output_dim;
        assert_eq!(
            layer.param_count(),
            TrainingParameters::Trainable(expected),
            "({}, {}) layer parameter count error",
            input_dim,
            output_dim
        );
    }
}

// Helper functions
fn calculate_mse(pred: &Tensor, target: &Tensor) -> f32 {
    let pred_2d = pred
        .as_standard_layout()
        .into_dimensionality::<ndarray::Ix2>()
        .unwrap();
    let target_2d = target
        .as_standard_layout()
        .into_dimensionality::<ndarray::Ix2>()
        .unwrap();

    let diff = &pred_2d - &target_2d;
    let squared_diff = diff.mapv(|x| x * x);
    squared_diff.mean().unwrap()
}