Closed hurri2000 closed 10 months ago
I think it should be : ... yi max(0, dist_p - m+) + (1-yi)max(0,m- - dist_n), in this way, the dist_p gets smaller and dist_n gets greater. please check.
Thanks for your great working.
Best regards,
The inner product is cosin similarity, so it is max(m+- dist_ p)
I see, thanks.
I think it should be : ... yi max(0, dist_p - m+) + (1-yi)max(0,m- - dist_n), in this way, the dist_p gets smaller and dist_n gets greater. please check.
Thanks for your great working.
Best regards,