. 2020 Aug 14;8:867. doi: 10.3389/fbioe.2020.00867

Table 2.

The algorithm details of attentional matching.

Algorithm1: Attentional matching
For each H′ from Multi-CNN: 1. Calculate label representation vector D; _{D = (W_gyg}+b) 2. The a_i Measures how informative each n-gram is for the i-th label.; $a_{i} = S o f t M a x ({H^{'}}^{T} D_{i}), i = 1, 2, 3 \dots, l_{g}$ 3. Calculate the weighted average v_i of the rows in H′ forming a vector representation of the clinic text for the i-th label; $v_{i} = a_{i} H^{'}$

Algorithm1: Attentional matching

For each H′ from Multi-CNN:
1. Calculate label representation vector D;
_{D = (W_gyg}+b)
2. The a_i Measures how informative each n-gram is for the i-th label.;

a_{i} = S o f t M a x ({H^{'}}^{T} D_{i}), i = 1, 2, 3 \dots, l_{g}

3. Calculate the weighted average v_i of the rows in H′ forming a vector representation of the clinic text for the i-th label;

v_{i} = a_{i} H^{'}