This site does not allow itself to be embedded. You can still read it on the original site — the toolbar below keeps your place in the directory.
Vector search: The problem, Part 4. Compressing the database vectors so a billion-vector index fits in a fraction of its raw memory. Scalar quantization (one byte per value), product quantization (the workhorse of every billion-scale vector index), how compressed vectors are searched without ever decompressing them, and how the codebook is kept fresh as data drifts.
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.