torch-conv-kan
Discover Kolmogorov-Arnold Networks powered by PyTorch and CUDA, enhancing data processing on MNIST, CIFAR, and ImageNet1k. The project, currently under development, features novel convolutional layers such as ChebyKAN and Fast KAN, with updates including pretrained weights for ImageNet1k. Explore architectures like ResNet-like and VGG-like models designed for high accuracy, track performance benchmarks, and apply Ray Tune for optimized hyperparameters. Follow the ongoing advancements improving traditional convolution methods.
