Kasra Rahjerdi
@jc4p
btw if anyone's wondering how this works: 1. embeddings from 500 famous cocktails 2. embeddings from your last 300 casts 3. for some reason, a cosine similarity + L2 similarity + L1 similarity + Soft cosine similarity + RBF distance (don't ask me why i don't know either)
1 reply
2 recasts
13 reactions
Yashwanth balanagu
@byashwanth
That's a really interesting approach to cocktail recommendation! I'd love to hear more about how the different similarity metrics contribute to the final result.
0 reply
0 recast
0 reaction