The field of spatial modeling and filtering is experiencing significant developments, with a notable shift towards incorporating neural networks and non-stationary models to improve accuracy and robustness. Researchers are exploring ways to enhance traditional methods, such as Geographically Weighted Regression and Kalman filters, by integrating concepts from deep learning and non-linear propagation models. These innovations enable better handling of complex spatial relationships, non-stationarity, and non-linearity, leading to improved performance in various applications, including navigation and regression tasks. Noteworthy papers include the proposal of a generalized Geographically Neural Network Weighted Regression framework, which leverages concepts from convolutional neural networks, recurrent neural networks, and transformers to capture spatial non-stationarity. Another significant contribution is the introduction of the Natural Gradient Gaussian Approximation filter, which outperforms popular Kalman filters in handling non-linearity and non-Gaussian systems. Additionally, a framework for non-stationary Gaussian processes with neural network parameters has been proposed, demonstrating improved accuracy and flexibility in modeling non-stationary data.