This paper introduces Convolutional Kolmogorov-Arnold Networks (Convolutional KANs), an innovative alternative to traditional Convolutional Neural Networks (CNNs) in computer vision. Convolutional KANs integrate non-linear activation functions from Kolmogorov-Arnold Networks (KANs) into convolutions, reducing the number of parameters while maintaining or improving accuracy. The authors validate the performance of Convolutional KANs against traditional architectures on the MNIST and Fashion-MNIST datasets, demonstrating that Convolutional KANs achieve similar accuracy with half the parameters. The paper also discusses the theoretical foundation of KANs, their architecture, and the motivation for using KANs. The experiments show that Convolutional KANs can maintain or slightly improve accuracy compared to CNNs while significantly reducing the number of parameters. The authors conclude that Convolutional KANs have the potential to be a promising alternative to CNNs, with future work focusing on optimizing and expanding their application to more complex datasets.This paper introduces Convolutional Kolmogorov-Arnold Networks (Convolutional KANs), an innovative alternative to traditional Convolutional Neural Networks (CNNs) in computer vision. Convolutional KANs integrate non-linear activation functions from Kolmogorov-Arnold Networks (KANs) into convolutions, reducing the number of parameters while maintaining or improving accuracy. The authors validate the performance of Convolutional KANs against traditional architectures on the MNIST and Fashion-MNIST datasets, demonstrating that Convolutional KANs achieve similar accuracy with half the parameters. The paper also discusses the theoretical foundation of KANs, their architecture, and the motivation for using KANs. The experiments show that Convolutional KANs can maintain or slightly improve accuracy compared to CNNs while significantly reducing the number of parameters. The authors conclude that Convolutional KANs have the potential to be a promising alternative to CNNs, with future work focusing on optimizing and expanding their application to more complex datasets.