In tests using standard image classification datasets (CIFAR-100, TinyImageNet), L p-Convolution significantly improved accuracy on both established models like AlexNet and modern architectures ...