AI Paper Atlas

Research map / Point clouds / Point cloud classification

On the Spectral Differences Between NTK and CNTK and Their Implications for Point Cloud Recognition

Yuanqu Mou, Chang Gou, Haiyang Bai, Jia Liu

ICLR 2026 · Point clouds › Point cloud classification · approach: Local feature aggregation networks

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
use 1D convolutional NTK instead of MLP NTK for point cloudsNTK, L=7 · ModelNet10 (xyz only)accuracy: 17.58 -> 76.65Improved
use 1D convolutional NTK instead of MLP NTK for point cloudsNTK, L=7 · ModelNet40accuracy: 11.10 -> 60.62Improved
use 2D convolutional NTK instead of MLP NTK for imagesNTK, L=7 · MNISTaccuracy: 94.00 -> 96.00Improved
use 2D convolutional NTK instead of MLP NTK for imagesNTK, L=7 · CIFAR10accuracy: 59.19 -> 76.79Improved
add MLP layers after convolutions (PointNTK vs 1dCNTK) (ablation)1dCNTK · ModelNet40accuracy: 60.62 -> 80.47Improved
add MLP layers after convolutions (PointNTK vs 1dCNTK) (ablation)1dCNTK · ModelNet10 (xyz+orientation, full train)accuracy: 91.96 -> 91.19No clear effect
use PointNTK kernel regression instead of PointNet with 100 training samplesPointNet · ModelNet40 (100 train samples)accuracy: 38.32 -> 45.71Improved
use PointNTK instead of PointNet on full training dataPointNet · ModelNet40accuracy: 88.71 -> 80.47Hurt
increase PointNTK depth from 7 to 15 layers (L) (ablation)PointNTK · ModelNet40accuracy: 80.47 -> 82.50Improved
replace max pooling and CE loss by average pooling and least squares (PointNetm) (ablation)PointNet · ModelNet40 (100 train samples)accuracy: 38.32 -> 5.17Hurt
add global average pooling to 1dCNTK (ablation)1dCNTK · ModelNet10 (xyz only), D=5accuracy: 44.38 -> 76.87Improved
add global average pooling to supervised 1dCNN (ablation)1dCNN · ModelNet10 (xyz only), D=5accuracy: 42.18 -> 39.54No 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