AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Hadamard Matrix Guided Online Hashing

Mingbao Lin, Rongrong Ji, Hong Liu, Xiaoshuai Sun, Chen Shen, Qi Tian

IJCV 2020 · Label-efficient and robust learning › Recognition and long-tail · approach: Learning from unlabeled and positive-unlabeled data · 51 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
replace linear regression on tanh-relaxed codes (HCOH) with perceptron bit classifiers, LSH-aligned Hadamard targets and model ensemblingHCOH · CIFAR-10mAP: average increase of 4.926%Improved
replace linear regression with perceptron classifiers on Hadamard targets (HMOH vs HCOH) at low code lengthHCOH · Places205 (8-bit)mAP@1000: increase of 108.163%Improved
replace linear regression with perceptron classifiers on Hadamard targets (HMOH vs HCOH) at 16 bitsHCOH · Places205 (16-bit)mAP@1000: 34.104%Improved
HMOH vs HCOH, averaged over bit lengthsHCOH · MNISTmAP: 15.595%Improved
HMOH vs best baseline (MIHash or BSODH), averaged over bit lengthsBSODH / MIHash · CIFAR-10mAP: average percentage of 7.478%Improved
increase code length from 64 to 128 bitsHMOH · CIFAR-10Precision@H2: 0.710 -> 0.734Improved
increase code length from 64 to 128 bitsHMOH · Places205Precision@H2: 0.223 -> 0.137Hurt
increase code length from 64 to 128 bitsHMOH · NUS-WIDEPrecision@H2: 0.474 -> 0.468No clear effect
use more training data (2K to 20K samples)HMOH · MNIST (48-bit)mAP: 0.780 -> 0.845Improved
add kernelization (RBF anchors) (ablation)HMOH · CIFAR-10 and NUS-WIDEmAPHurt
add ensembling of models learned in every round (ablation)HMOH · benchmarksmAPImproved
more training data for low bit codes (8, 16, 32 bits)HMOH · NUS-WIDEmAPHurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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