PulseAugur
EN
LIVE 10:08:43

New SKANs offer parameter-efficient alternative to KANs

By PulseAugur Editorial · [3 sources] ·

Researchers have introduced Structural Kolmogorov-Arnold Convolutions (SKANs) as a more parameter-efficient alternative to existing Convolutional Kolmogorov-Arnold Networks (KANs). The new approach repositions learnable functions from individual kernel entries to the overall structure of the convolution, either acting on pixel values or filter shapes. Experiments show that SKAN variants, specifically RF-KAN and SV-KAN, achieve competitive accuracy on CIFAR-10 and CIFAR-100 datasets with significantly fewer parameters than traditional KANs. AI

IMPACT This research could lead to more efficient deep learning models by reducing parameter count while maintaining performance.

RANK_REASON The cluster contains a research paper detailing a new model architecture.

Read on arXiv cs.AI →

AI-generated summary · Google Gemini · from 3 sources. How we write summaries →

New SKANs offer parameter-efficient alternative to KANs

COVERAGE [3]

  1. arXiv cs.AI TIER_1 English(EN) · Stefano Mereu, Oleksandr Kuznetsov, Gabriele Marchello, Alessandro Galdelli, Emanuele Frontoni, Adriano Mancini, Ferdinando Cannella ·

    Structural Kolmogorov-Arnold Convolutions: Learnable Function on the Values or the Filter Shape as Parameter-Efficient Alternative to Per-Edge Convolutional KANs

    arXiv:2606.24371v1 Announce Type: cross Abstract: Convolutional Kolmogorov--Arnold Networks (KANs) replace the fixed weights of a convolutional kernel with learnable univariate functions. The dominant formulation attaches one such function to every kernel entry and lets it act on…

  2. arXiv cs.AI TIER_1 English(EN) · Ferdinando Cannella ·

    Structural Kolmogorov-Arnold Convolutions: Learnable Function on the Values or the Filter Shape as Parameter-Efficient Alternative to Per-Edge Convolutional KANs

    Convolutional Kolmogorov--Arnold Networks (KANs) replace the fixed weights of a convolutional kernel with learnable univariate functions. The dominant formulation attaches one such function to every kernel entry and lets it act on pixel values, expressive but parameter-heavy and …

  3. Hugging Face Daily Papers TIER_1 English(EN) ·

    Structural Kolmogorov-Arnold Convolutions: Learnable Function on the Values or the Filter Shape as Parameter-Efficient Alternative to Per-Edge Convolutional KANs

    Convolutional Kolmogorov--Arnold Networks (KANs) replace the fixed weights of a convolutional kernel with learnable univariate functions. The dominant formulation attaches one such function to every kernel entry and lets it act on pixel values, expressive but parameter-heavy and …

RELATED ENTITIES

RELATED TOPICS