HuangYG123
commented
4 years ago cos_theta is the cosine similarity, and the shape is BxC. cos_theta_m is the target cosine similarity, and the shape is Bx1. The shape of mask is BxC. Why cos_theta is always larger than cos_thata_m?
zhushumin
In this loss, you choose the hard_example by "mask = cos_theta > cos_theta_m; " and " hard_example = cos_theta[mask]". However the cos_theta is always larger than cos_thata_m. This code is useless and it does not match your paper.
if cos(θyi + m) ≥ cosθj then N(t,cosθj) = cosθj; else N(t,cosθj) = (t(k) + cosθj)cosθj ;