AI Paper Atlas

Research map / Label-efficient and robust learning / Domain adaptation and generalization

Reducing Distributional Uncertainty by Mutual Information Maximisation and Transferable Feature Learning

Jian Gao, Yang Hua, Guosheng Hu, Chi Wang, Neil M. Robertson

ECCV 2020 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Statistical alignment and hypothesis testing · 17 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 MI maximisation (MIM) to transferable feature learning (ablation)TFL · Office-Home single-to-single UDA (ResNet-50)mean accuracy (%): 63.4 -> 64.7Improved
add transferable shape feature learning (TFL) to MI maximisation (ablation)MIM · Office-Home single-to-single UDA (ResNet-50)mean accuracy (%): 63.1 -> 64.7Improved
add MI maximisation (MIM) to transferable feature learning (ablation)TFL · DomainNet single-to-single (15 difficult tasks)mean accuracy (%): 10.8 -> 12.6Improved
add MI maximisation (MIM) to transferable feature learning (ablation)TFL · DomainNet single-to-single (all 30 tasks)all mean accuracy (%): 27.3 -> 28.1Improved
add MEC lossMIMTFL · Office-Home single-to-single UDA (ResNet-50)mean accuracy (%): 64.7 -> 67.2Improved
train with target data (UDA) instead of source onlyMIMTFL src · Office-Home single-to-single UDA (ResNet-50)mean accuracy (%): 63.1 -> 64.7Improved
use MIMTFL instead of source-only trainingSource only · DomainNet single-to-single (15 difficult tasks)mean accuracy (%): 10.8 -> 12.6Improved
use ADDA instead of source-only trainingSource only · DomainNet single-to-single (15 difficult tasks)mean accuracy (%): 10.8 -> 9.7Hurt
use MCD instead of source-only trainingSource only · DomainNet single-to-single (15 difficult tasks)mean accuracy (%): 10.8 -> 9.2Hurt
use SE instead of source-only trainingSource only · Office-Home multi-to-multi UDAmean accuracy (%): 70.2 -> 65.6Hurt
use MIMTFL+MEC instead of source-only trainingSource only · Office-Home APR->C multi-to-singleaccuracy (%): 58.5 -> 64.3Improved
use shape-learning MIMTFL instead of jigsaw-puzzle JiGenJiGen · Office-Home multi-to-single DG (ResNet-18)mean accuracy (%): 61.2 -> 63.0Improved

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