Current state: ["Under Discussion"]
...
This API does range search for index, it returns all unsorted results with distance "better than radius" (for IP: > radius; for others: < radius).
| PROTO | virtual DatasetPtr
QueryByRange(const DatasetPtr& dataset, const Config& config, const faiss::BitsetView bitset) |
INPUT | Dataset {
knowhere::meta::TENSOR: - // query data
knowhere::meta::ROWS: - // rows of queries
knowhere::meta::DIM: - // dimension
} Config { knowhere::meta::RADIUS: - // radius for range search } |
OUTPUT | Dataset {
knowhere::meta::IDS: - // result IDs with length LIMS[nq]
knowhere::meta::DISTANCE: - // result DISTANCES with length LIMS[nq]
knowhere::meta::LIMS: - // result offset prefix sum with length nq + 1
} |
...
This API does range search for no-index dataset, it returns all unsorted results with distance "better than radius" (for IP: > radius; for others: < radius).
| PROTO | static DatasetPtr
RangeSearch(const DatasetPtr base_dataset,
const DatasetPtr query_dataset,
const Config& config,
const faiss::BitsetView bitset); |
INPUT | Dataset {
knowhere::meta::TENSOR: - // base data
knowhere::meta::ROWS: - // rows of base data
knowhere::meta::DIM: - // dimension
} Dataset {
knowhere::meta::TENSOR: - // query data
knowhere::meta::ROWS: - // rows of queries
knowhere::meta::DIM: - // dimension
} Config { knowhere::meta::RADIUS: - // radius for range search } |
OUTPUT | Dataset {
knowhere::meta::IDS: - // result IDs with length LIMS[nq]
knowhere::meta::DISTANCE: - // result DISTANCES with length LIMS[nq]
knowhere::meta::LIMS: - // result offset prefix sum with length nq + 1
} |
...
- use sift1M/glove200 dataset to test range search (radius_low_bound = -1.0, radius_high_bound = 1000000max_float), we expect to get identical result results as search
Rejected Alternatives(optional)
...
- Knowhere Range Search Algorithm Introduction Introductio (https://zilliverse.feishu.cn/docx/doxcn1Kuj75rTiv6fPVburRF37c?from=from_copylink)