I have the following algorithm that I'm trying to implement:
$$H_{am} \leftarrow H_{am}\frac{ \sum_{i}W_{ia}X_{im}/(W \cdot H)_{im}}{\sum_{k}W_{ka} }$$
Where for example $X$ is a of size 5 x 10, $H$ is of size 2 x 10 and $W$ is of size 5 x 2.
What is the meaning of the subscript symbols $_{a}$ and $_{m}$ within the summation notation and outside of it? I thought initially the $_{a}$ and $_{m}$ were exactly the dimensions of $H$ (2 x 10) so in the case of the $W$ denominator this would be every row of $W$ iteratively $_{k}:$(1-5) of column $_{a}$(2), however following the same logic and updating the final element of $H_{am}$ with the result seems bizarre and doesn't work.
I also thought that $_a$ and $_{m}$ may have meant all the rows/columns respectively but that doesn't really make much sense with the $_{i}$/$_{k}$ and even if it was a selected vector the element wise multiplication would be a mismatch of sizes e.g $W$ and $X$.
Any suggestions would be most welcome.