AI Paper Atlas

Research map / Architectures and efficiency / Quantization

Unsupervised Neural Quantization for Compressed-Domain Similarity Search

Stanislav Morozov, Artem Babenko

ICCV 2019 · Architectures and efficiency › Quantization · approach: Post-training weight quantization · 25 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
use UNQ instead of LSQLSQ · BigANN1M (8 bytes)Recall@1: 29.2 -> 34.6Improved
use UNQ instead of LSQLSQ · Deep1M (16 bytes)Recall@1: 41.1 -> 47.9Improved
add learned-decoder reranking to shallow LSQLSQ · BigANN1M (8 bytes)Recall@1: 29.2 -> 30.3No clear effect
use UNQ instead of Catalyst + LatticeCatalyst + Lattice · Deep1B (8 bytes)Recall@1: 16.8 -> 14.5Hurt
use UNQ instead of Catalyst + LatticeCatalyst + Lattice · BigANN1B (16 bytes)Recall@1: 31.1 -> 38.3Improved
add decoder reranking of top candidates (ablation)UNQ · BigANN1M (8 bytes)Recall@1: 25.0 -> 34.6Improved
rerank exhaustively with d1 instead of top-500 candidates (ablation)UNQ · BigANN1M (8 bytes)Recall@1: 34.6 -> 34.6No clear effect
add triplet loss on the learned-space distance d2 (ablation)UNQ · BigANN1M (8 bytes)Recall@100: 95.7 -> 99.0Improved
add triplet loss on the learned-space distance d2 (ablation)UNQ · BigANN1M (8 bytes)Recall@1: 35.5 -> 34.6Hurt
use triplet loss only (no reconstruction term) instead of combined objective (ablation)UNQ · BigANN1M (8 bytes)Recall@1: 27.9 -> 34.6Hurt
add the CV^2 balance regularizer (ablation)UNQ · BigANN1M (8 bytes)Recall@1: 31.0 -> 34.6Improved
use hard (argmax straight-through) Gumbel-Softmax instead of soft (ablation)UNQ · BigANN1M (8 bytes)Recall@1: 33.8 -> 34.6Improved

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