gradients 0.3.4

An OpenCL, CUDA and CPU based Deep Learning 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

use custos::CDatatype;
use custos_math::Matrix;

pub fn find_idxs<T: Copy + Default + PartialEq>(
    search_for: &Matrix<T>,
    search_with: &Matrix<T>,
) -> Vec<usize> {
    let rows = search_for.rows();
    let search_for = search_for.read();
    let search_with = search_with.read();
    purpur::utils::find_idxs(rows, &search_for, &search_with)
}

pub fn correct_classes<T: CDatatype>(targets: &[usize], search_for: &Matrix<T>) -> usize {
    let search_with = search_for.max_cols();
    let idxs = find_idxs(&search_for, &search_with);
    let mut correct = 0;
    for (idx, correct_idx) in idxs.iter().zip(targets) {
        if idx == correct_idx {
            correct += 1;
        }
    }
    correct
}