179 questions
No questions match those filters.
← All questions
Retrieval & EmbeddingsHard· HNSW· IVF-PQ· Memory Optimization
One team wants HNSW for a 200 M-chunk index, another wants IVF-PQ, and the RAM budget just got cut. Adjudicate.
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 plansAt 200 million vectors of 768 dimensions, raw float32 storage is 614 GB, and HNSW graph links add significant overhead. The choice between IVF and HNSW is secondary to the need for quantization. If RAM is free and the corpus is static, HNSW offers better recall-per-millisecond. However, IVF is often preferred if the budget forces aggressive compression or if rebuild costs are a primary concern.