AI Paper Atlas

Research map / Architectures and efficiency / CNN design

Topological Convolutional Layers for Deep Learning

Ephy R. Love, Benjamin Filippenko, Vasileios Maroulas, Gunnar Carlsson

JMLR 2023 · Architectures and efficiency › CNN design · approach: Filter design and trained-filter analysis

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 learned first-layer filters with fixed Klein bottle filters and Klein-bottle-sparse connections (TCNN)12-layer ResNet · UCF-101test accuracy after 100 epochs: 55% -> 70%Improved
use M-F topological first layer instead of learned NOL layerResNet trained on KTH · KTH to Weizmanngeneralization accuracy: about 52% -> about 65%Improved
use KF + KOL topological layers instead of NOL + NOL2-layer CNN · MNIST to SVHNgeneralization accuracy: 10% -> 30%Improved
use KF + KOL topological layers instead of NOL + NOL2-layer CNN · SVHN to MNISTgeneralization accuracy: over 60% accuracy for KF + KOLImproved
add Klein bottle filter layer (KF) with KOL or COL2-layer CNN · MNIST, SVHN, USPS trained on 1,000 imagestest accuracyImproved
add topological layers under class-correlated Gaussian noise in training or test data2-layer CNN · noisy MNISTtest accuracyImproved
use Klein bottle filters instead of an arbitrary Gabor filter family in the fixed filter layer (ablation)KF + NOL network · 3 digit datasets and 2 cats-dogs datasetstest accuracyImproved
use Klein bottle connectivity in KOL instead of randomly permuted connections with same sparsity (ablation)KF + KOL · CIFAR-10accuracy over trainingImproved
add a pooling layer (ablation)CNN and TCNN for digit generalization · MNIST/SVHN generalizationgeneralization accuracyNo clear effect
use TCNN instead of CNN on the simpler KTH dataset12-layer ResNet · KTHtest accuracyImproved

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