AI Paper Atlas

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

Direct Hashing Without Pseudo-Labels

Feng Zheng, Heng Huang

AAAI 2018 · Label-efficient and robust learning › Recognition and long-tail · approach: Learning from unlabeled and positive-unlabeled data · 1 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 W-shape loss instead of ITQ's squared quantization approachITQ · Toy setaverage quantization loss: 1.599 -> 0.783Improved
Use W-shape loss instead of IsoHashIsoHash · Toy setaverage quantization loss: 1.59 -> 0.783Improved
Use WSL instead of ITQ at 16 bitsITQ · CIFAR-10MAP: 0.167 -> 0.183Improved
Use WSL instead of ITQ at 64 bitsITQ · CIFAR-10MAP: 0.179 -> 0.227Improved
Use WSL instead of ITQ at 16 bitsITQ · MNISTMAP: 0.410 -> 0.434Improved
Use WSL instead of ITQ at 64 bitsITQ · MNISTMAP: 0.456 -> 0.522Improved
Use WSL instead of SpH at 64 bitsSpH · MNISTMAP: 0.356 -> 0.522Improved
Increase hash code length from 16 to 64 bits (ablation)WSL · CIFAR-10MAP: 0.183 -> 0.227Improved
Increase hash code length from 16 to 64 bitsPCAH · MNISTMAP: 0.276 -> 0.212Hurt
Set mu too small (ablation)WSL · Embedding studyprojection norm / embedded valuesHurt
Set mu too large (ablation)WSL · Embedding studyretrieval performanceHurt
Use linear WSL instead of nonlinear AGHAGH · SIFT1Mrecall@RImproved

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