AI Paper Atlas

Research map / Optimization / Optimizers

Accelerate Learning of Deep Hashing With Gradient Attention

Long-Kai Huang, Jianda Chen, Sinno Jialin Pan

ICCV 2019 · Optimization › Optimizers · approach: Constrained and non-backprop training · 29 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 gradient attention network instead of random weight on one gradient (B1) (ablation)GAH · ImageNet (64 bits)MAP: 0.6483 -> 0.6821Improved
add gradient attention instead of random weights (B2) (ablation)GAH · ImageNet (64 bits)MAP: 0.6536 -> 0.6821Improved
add gradient attention instead of weight 1 to larger gradient (B3) (ablation)GAH · ImageNet (64 bits)MAP: 0.5670 -> 0.6821Improved
add gradient attention instead of zeroing smaller gradient (B4) (ablation)GAH · ImageNet (64 bits)MAP: 0.6577 -> 0.6821Improved
add gradient attention network to HashNet-like loss (GAH vs HashNet)deep hashing · CIFAR-10 (16 bits)MAP: 0.6896 -> 0.6985Improved
add gradient attention (GAH vs HashNet)deep hashing · ImageNet (16 bits)MAP: 0.4633 -> 0.5034Improved
add gradient attention (GAH vs GH)deep hashing · ImageNet (64 bits)MAP: 0.4893 -> 0.6821Improved
add gradient attention (GAH vs GH) in unsupervised settingdeep hashing · CIFAR-10 unsupervised (16 bits)MAP@5000: 0.4114 -> 0.4796Improved
add gradient attention (GAH vs GH) in unsupervised settingdeep hashing · CIFAR-10 unsupervised (64 bits)MAP@5000: 0.5034 -> 0.5165Improved
add gradient attention to reduce bit pairs in dilemma (GAH0 vs GAH) (ablation)GAH · CIFAR-10bit pairs in dilemma: 13.1% to 61.9%Improved
add gradient attention to reduce bit pairs in dilemma (ablation)GAH · NUS-WIDEbit pairs in dilemma: at least 26.1%Improved
add gradient attention to reduce bit pairs in dilemma (ablation)GAH · ImageNetbit pairs in dilemma: at least 10.8%Improved

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