384 is a tad small. How does it perform on 1k/2k/4k..?
Just brute force your embeddings
11–15 of 15 posts
Re: Just brute force your embeddings
#12Similarly, COST, configuration that outperforms a single thread:
https://www.usenix.org/system/files/conference/hotos15/hotos...
Re: Just brute force your embeddings
#13I don't understand what he means by "brute force your embedding". Does this mean "don't calculate your embedding ahead of time, just generate them as you need them?
Re: Just brute force your embeddings
#14I don't understand what he means by "brute force your embedding". Does this mean "don't calculate your embedding ahead of time, just generate them as you need them?
Looking at the linked article.
He is suggesting that you don’t need a vector database.
Re: Just brute force your embeddings
#15This works especially well if your embedding model was trained to perform well with quantized embeddings. Binary + hamming distance = incredibly fast. This post is from 2024 but I wrote about using this technique in https://emschwartz.me/binary-vector-embeddings-are-so-cool/
I remember similar observations for an earlier use case in computer vision, loop closure and place recognition for visual SLAM algorithms. In this case the goal was to find a needle (or needles) in a big haystack of visual descriptors (in some sense, proto-embeddings for small image patches or in some cases, whole images). Several approaches used hierarchical data structures for the NN search such as k-mean trees. But linear search - especially with binary descriptors, also became popular as a fast and simple alternative.