AI Paper Atlas

Research map / Architectures and efficiency / CNN design

Diverse feature visualizations reveal invariances in early layers of deep neural networks

Santiago A. Cadena, Marissa A. Weis, Leon A. Gatys, Matthias Bethge, Alexander S. Ecker

ECCV 2018 · Architectures and efficiency › CNN design · approach: Biologically inspired CNN explanation · 25 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 diversity term (lambda>0) to activation maximization (ablation)Hubel & Wiesel complex cell model · phase coverage of pre-imagescoverage of invariant subspace: Four out of six solutions at lambda=0 are the globally best imageImproved
increase diversity weight lambda too far (lambda > 2) (ablation)complex cell model · activation levelaverage activationHurt
train VGG-19 on ImageNet instead of random weights (ablation)VGG-19 conv3_1 · invariance to diverse pre-imagesimage diversity at 90% activation: at least two and up to 100-foldImproved
use ResNet-50 instead of VGG-19early layers · phase invarianceshift invariance indexHurt
use ResNet-50 instead of VGG-19early layers · tolerance to local deformationslinear combination indexNo clear effect
use minimum distance rather than average distance in diversity term (ablation)diversity objective · Energy model coveragephase coverageImproved
go to higher layers within conv3VGG-19 · conv3 layersshift invarianceHurt

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