Timeline for Fastest way for 1 vs all lookup on embeddings
Current License: CC BY-SA 4.0
8 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Mar 14, 2023 at 7:16 | comment | added | noe | @Avv Faiss is a database for vectors. There, you store vector representations of text, of images, or whatever data you have (i.e. embeddings). Then you can search by vector similarity based on another query vector. This allows you to implement for instance search engines based on the similarity of the content. Please consider upvoting this answer if it was useful to you. You can create a new question on this site if you have further doubts. | |
| Mar 14, 2023 at 2:42 | comment | added | Avv | Thanks. May I know the difference between Faiss index and database index please? | |
| Mar 16, 2020 at 19:27 | history | edited | noe | CC BY-SA 4.0 |
added 159 characters in body
|
| Mar 16, 2020 at 16:36 | comment | added | noe |
The expression of the cosine distance is the inner product of the normalized vectors. IndexFlatIP does not compute the cosine distance, but the inner product. That's why you need to normalize the vectors.
|
|
| Mar 16, 2020 at 16:31 | vote | accept | Isbister | ||
| Mar 16, 2020 at 16:31 | comment | added | Isbister | Wow, this was exactly what I was looking for. Do you know why you have to normalize the vectors first? To my basic knowledge, cosine similarity considers the angles of the high dimensional vectors, should they not be the same regardless normalization? | |
| Mar 16, 2020 at 16:30 | history | edited | noe | CC BY-SA 4.0 |
added 269 characters in body
|
| Mar 16, 2020 at 16:15 | history | answered | noe | CC BY-SA 4.0 |