AI Paper Atlas

Research map / Reinforcement learning / Imitation learning

Generative Adversarial Regularized Mutual Information Policy Gradient Framework for Automatic Diagnosis

Yuan Xia, Jingbo Zhou, Zhenhui Shi, Chao Lu, Haifeng Huang

AAAI 2020 · Reinforcement learning › Imitation learning · approach: Imitation combined with RL · 67 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 mutual information reward to vanilla policy gradient (ablation)policy gradient · Dxydiagnosis accuracy: 0.731 -> 0.749Improved
add mutual information reward to vanilla policy gradient (ablation)policy gradient · Dxyaverage turns: 3.11 -> 2.76Improved
add GAN discriminator reward to vanilla policy gradient (ablation)policy gradient · Dxydiagnosis accuracy: 0.731 -> 0.758Improved
add GAN discriminator reward to vanilla policy gradient (ablation)policy gradient · Dxyaverage turns: 3.11 -> 3.4Hurt
combine MI and GAN rewards versus MI only (ablation)PG + MI · Dxydiagnosis accuracy: 0.749 -> 0.769Improved
combine MI and GAN rewards versus GAN only (ablation)PG + GAN · Dxydiagnosis accuracy: 0.758 -> 0.769Improved
combine MI and GAN rewards versus GAN only (ablation)PG + GAN · Dxyaverage turns: 3.4 -> 2.68Improved
policy gradient with fixed reward over MLE pretrained LSTM (ablation)pretrained LSTM · Dxydiagnosis accuracy: 0.643 -> 0.731Improved
add MI and GAN rewards to vanilla policy gradient (ablation)policy gradient · Dxymatch rate: 0.212 -> 0.179Hurt

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