Kolmogorov-Arnold Networks for Time Series Analysis and Forecasting: A Survey

Overview & Synopsis

Kolmogorov-Arnold Networks (KANs) emerged as a powerful alternative to Multi-Layer Perceptron (MLP). This approach leverages adaptive activation functions to enhance function approximation capabilities. Recently, several variants of KANs have been developed for time series analysis and forecasting. This survey provides a comprehensive overview of these models and their frameworks. We examine their architectures, foundations and conclusion in time series forecasting tasks. This survey analyzes how different models integrate KAN components with established time series architectures, such as RNNs, LSTMs, and Transformers. Additionally, we highlight the advantages, challenges, and potential future directions in this rapidly evolving field.