Hi,
I read your ESM-2 paper and I'm curious about the details of outerdifference.
the outer product is calculated as: (c, 1) dot (1, c) → (c, c)
So I wonder if outer difference is calculated as: (c, 1) - (1, c) → (c, c) by pytorch broadcast
yes that sounds right.
It is evolved from the "outer concatenation" described in Wang et al 2017 Accurate De Novo Prediction of Protein Contact Map by Ultra-Deep Learning Model
Hi, I read your ESM-2 paper and I'm curious about the details of outerdifference. the outer product is calculated as: (c, 1) dot (1, c) → (c, c) So I wonder if outer difference is calculated as: (c, 1) - (1, c) → (c, c) by pytorch broadcast
Thanks in advance!