179 questions
No questions match those filters.
← All questions
Retrieval & EmbeddingsHard· Product Quantization· Computational Complexity
Why use a product codebook (multiple small codebooks) i...
This is one of the questions in the full AI/ML interview bank. Pro unlocks all 1789 questions; Premium includes the same bank plus the highest daily Practice limit.
See plansA single codebook with 2^32 centroids is theoretically more expressive but practically impossible due to the massive training data and memory required for centroids. Product Quantization (PQ) breaks the space into smaller subspaces, which is far more efficient. The trade-off is that PQ assumes boundaries are axis-parallel in the subspace decomposition. Techniques like Optimized Product Quantization (OPQ) can mitigate this by rotating the space to better align with the axes.