native_neural_network 0.1.6

Lib no_std Rust for native neural network (.rnn)
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

#[derive(Clone, Copy, Debug, PartialEq, Eq)]
pub enum GradientError {
    Empty,
    InvalidThreshold,
}

pub fn l2_norm(values: &[f32]) -> Result<f32, GradientError> {
    if values.is_empty() {
        return Err(GradientError::Empty);
    }
    let mut sum = 0.0f32;
    for &v in values {
        sum += v * v;
    }
    Ok(crate::math::sqrtf(sum))
}

pub fn clip_by_global_norm(values: &mut [f32], max_norm: f32) -> Result<f32, GradientError> {
    if values.is_empty() {
        return Err(GradientError::Empty);
    }
    if !max_norm.is_finite() || max_norm <= 0.0 {
        return Err(GradientError::InvalidThreshold);
    }

    let norm = l2_norm(values)?;
    if norm > max_norm {
        let scale = max_norm / norm;
        for v in values.iter_mut() {
            *v *= scale;
        }
    }
    Ok(norm)
}

pub fn all_finite(values: &[f32]) -> bool {
    values.iter().all(|v| v.is_finite())
}