AI Paper Atlas

Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning

Evolving Semantic Prototype Improves Generative Zero-Shot Learning

Shiming Chen, Wenjin Hou, Ziming Hong, Xiaohan Ding, Yibing Song, Xinge You, Tongliang Liu, Kun Zhang

ICML 2023 · Label-efficient and robust learning › Zero-/few-shot and meta-learning · approach: Generative feature synthesis for zero-shot · 4 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 DSP to f-VAEGANf-VAEGAN · CUB (GZSL)H: 52.9 -> 67.4Improved
Add DSP to f-VAEGANf-VAEGAN · SUN (GZSL)H: 41.3 -> 48.1Improved
Add DSP to f-VAEGANf-VAEGAN · AWA2 (GZSL)H: 63.5 -> 74.2Improved
Add DSP to CLSWGAN, TF-VAEGAN, or FREEGenerative ZSL models · CUB, SUN, and AWA2 (GZSL)H: 49.7 -> 56.9; 39.4 -> 45.5; 59.6 -> 70.7; 57.5 -> 62.8; 42.1 -> 51.7; 66.3 -> 74.8; 57.7 -> 64.6; 41.7 -> 51.0; 67.1 -> Improved
Add semantic cycle consistency loss (ablation)DSP with f-VAEGAN · CUB and SUN (GZSL)H: 61.9 -> 67.4; 46.0 -> 48.1Improved
Add semantic reconstruction loss (ablation)DSP with f-VAEGAN · CUB and SUN (GZSL)H: 65.5 -> 67.4; 45.4 -> 48.1Improved
Add visual-to-semantic loss supervising prototype evolution (ablation)DSP with f-VAEGAN · CUB and SUN (GZSL)H: 59.9 -> 67.4; 42.7 -> 48.1Improved
Use smooth moving-average prototype updates (ablation)DSP with f-VAEGAN · CUB and SUN (GZSL)H: 54.8 -> 67.4; 42.1 -> 48.1Improved
Concatenate evolved prototypes with visual features (ablation)DSP with f-VAEGAN · CUB and SUN (GZSL)H: 53.1 -> 67.4; 42.5 -> 48.1Improved
Remove semantic reconstruction loss (ablation)DSP with f-VAEGAN · CUB and SUN (GZSL)H: 67.4 -> 65.5; 48.1 -> 45.4Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 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