BSRBF-KAN: A combination of B-splines and Radial Main Capabilities in Kolmogorov-Arnold Networks
Authors: Hoang-Thang Ta
Abstract: On this paper, we introduce BSRBF-KAN, a Kolmogorov Arnold Group (KAN) that mixes Bsplines and radial basis options (RBFs) to swimsuit enter vectors in info teaching. We stock out experiments with BSRBF-KAN, MLP, and completely different well-liked KANs, along with EfficientKAN, FastKAN, FasterKAN, and GottliebKAN over the MNIST dataset. BSRBF-KAN reveals stability in 5 teaching situations with a aggressive widespread accuracy of 97.55% and obtains convergence larger than completely different networks. We anticipate BSRBF-KAN can open many mixtures of mathematical options to design KANs. Our repo is publicly on the market at: https://github.com/hoangthangta/BSRBF-KAN.
