Showing page 1 of 1 pages of list content

  1. Kolmogorov-Arnold Networks: Key Developments and Uses

    This article has 1 author:
    1. Bochra Hadj Kilani
    This article has no evaluationsAppears in 1 listLatest version

    bochra hadj kilani

    Kolmogorov-Arnold Networks (KANs) have emerged as a promising advancement in the field of neural networks, offering enhanced interpretability, efficiency, and adaptability. This literature review explores various applications and developments of KANs. Temporal KANs (TKANs) combine the strengths of Recurrent Neural Networks (RNNs) and KANs for improved multi-step time series forecasting. DeepOKAN employs Gaussian radial basis functions (RBFs) in lieu of B-splines for computational mechanics, resulting in a notable acceleration and enhancement in efficiency. Wav-KANs utilize wavelet functions for data analysis, offering a balance between detail and overview. KANs have also been utilized in image classification, integrating with pre-trained Convolutional Neural Network (CNN) models for remote sensing scene classification tasks. The Variational Quantum Kolmogorov-Arnold Network (VQKAN) implements a quantum version of KAN on a quantum circuit, optimizing synaptic connection weights for improved efficiency. KANs have also been applied for Explainable Natural Language Processing (NLP), forming continuous word embeddings based on the meaning profiles of words. The review also mentions UKAN and IKAN, These advances underscore the versatility and potential of KANs in enhancing various machine learning applications. Future research will likely concentrate on further refinement of these architectures and exploring their suitability to diverse datasets and tasks.