|
Mr. Fasthorse@mrfasthorse.bsky.social |
I hope they're not matching like-to-like, not everyone matches with their identical self. The mapping of types to types would actually be kinda interesting but is probably a trade secret
1 replies 0 reposts 1 likes
|
Stephen Webb
@stephenwebb.bsky.social
[ View ] |
i would have a transformer layer that produces the feature vector for each profile separately, then toss the concatenated feature vectors into a feed forward binary classifier based on user feedback of how their match went. i think this would work very badly for dating, but this is what i would do.
1 replies 0 reposts 0 likes