AI Paper Atlas

Research map / Image generation / GANs

Predictive Adversarial Learning from Positive and Unlabeled Data

Wenpeng Hu, Ran Le, Bing Liu, Feng Ji, Jinwen Ma, Dongyan Zhao, Rui Yan

AAAI 2021 · Image generation › GANs · approach: Stabilized GAN 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
replace the GAN generator with a PU classifier and use PAN's KL-divergence adversarial objectivePU learning · YELP, RT, IMDB, 20News, MNIST, CIFAR10accuracy: 83.3 -> 83.6; 59.6 -> 64.1; 74.6 -> 78.8; 78.1 -> 81.0; 95.6 -> 96.4; 89.0 -> 89.7Improved
replace the GAN generator with a PU classifier and use PAN's KL-divergence adversarial objectivePU learning · YELPF-score: 83.7 -> 83.5Hurt
replace the GAN generator with a PU classifier and use PAN's KL-divergence adversarial objectivePU learning · RT, IMDB, 20News, MNIST, CIFAR10F-score: 66.6 -> 66.6; 76.2 -> 77.1; 78.5 -> 81.1; 95.6 -> 96.5; 86.6 -> 87.2Mixed
add term III to PAN (ablation)PAN · YELPaccuracy: 80.67 -> 83.56Improved
add term III to PAN (ablation)PAN · 20Newsaccuracy: 72.63 -> 81.00Improved
add term III to PAN (ablation)PAN · MNIST and CIFAR10accuracy: 96.30 -> 96.42; 89.38 -> 89.70Improved
add term III to PAN (ablation)PAN · YELP, 20News, MNIST, CIFAR10F-score: 81.86 -> 83.45; 73.59 -> 81.06; 96.27 -> 96.51; 86.68 -> 87.22Improved
set the positive training-data ratio to 1% rather than 10% (ablation)PAN · MNISTaccuracy: 96.42 -> 90.30Hurt
set the positive training-data ratio to 1% rather than 10% (ablation)PAN · CIFAR10accuracy: 90.77 -> 86.10Hurt
use λ=0.01 rather than λ=0.0001 (ablation)PAN · MNISTaccuracy: 90.30 -> 82.70Hurt
use λ=0.001 rather than λ=0.0001 (ablation)PAN · MNISTaccuracy: 88.90 -> 90.30No clear effect

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