AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Revisiting the Inverted Indices for Billion-Scale Approximate Nearest Neighbors

Dmitry Baranchuk, Artem Babenko, Yury Malkov

ECCV 2018 · Architectures and efficiency › RNNs and normalization · approach: Learned embeddings for retrieval · 104 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add grouping (L=64 subcentroids) to the inverted index (ablation)Inverted index K=2^20 · DEEP1Baverage distance to closest centroid: 0.282 -> 0.255Improved
add grouping to the inverted multi-index (ablation)IMI K=2^10 · DEEP1Baverage distance to closest centroid: 0.415 -> 0.385Improved
increase the inverted index codebook from K=2^18 to 2^22Inverted index · DEEP1Baverage distance to closest centroid: 0.315 -> 0.259Improved
increase the IMI codebook from K=2^13 to 2^15Inverted multi-index · DEEP1Baverage distance to closest centroid: 0.345 -> 0.305Improved
increase subcentroids per region from L=32 to 64 (ablation)IVFOADC+Grouping+Pruning · DEEP1B (16-byte codes)Recall@1: 0.417 -> 0.433Improved
increase subcentroids per region from L=64 to 128 (ablation)IVFOADC+Grouping+Pruning · DEEP1B (16-byte codes)Recall@1: 0.433 -> 0.441Improved
increase subcentroids per region from L=64 to 128 (runtime) (ablation)IVFOADC+Grouping+Pruning · DEEP1B (16-byte codes)search time (ms): 1.28 -> 1.48Hurt
use inverted index with large codebook, grouping and pruning instead of inverted multi-indexIVFOADC+G+P vs O-Multi-D-OADC · DEEP1B (16-byte codes)Recall@1: 0.397 -> 0.452Improved
use inverted index with grouping and pruning instead of inverted multi-indexIVFOADC+G+P vs O-Multi-D-OADC · SIFT1B (16-byte codes)Recall@1: 0.360 -> 0.405Improved
use inverted index with grouping and pruning instead of Multi-LOPQIVFOADC+G+P vs Multi-LOPQ · SIFT1B (16-byte codes)Recall@1: 0.454 -> 0.405Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy