The field of deep learning is moving towards incorporating equivariant principles to improve the robustness and expressiveness of neural networks. Recent works have focused on developing group equivariant convolutional networks, categorical equivariant deep learning, and shift-equivariant complex-valued convolutional neural networks. These advancements have shown promising results in various applications, including pathloss estimation, wireless communication, and computer vision. Notably, the development of datasets such as RadioMapMotion has enabled the evaluation of spatio-temporal radio environment prediction methods. The use of data augmentation techniques has also been explored to improve the reverse-engineering of neural network weights. Additionally, the analysis of dynamic sampling networks has led to a better understanding of their properties and training dynamics. Some particularly noteworthy papers include: RadioMapMotion, which introduces a large-scale public dataset for spatio-temporal radio environment prediction, and Group Equivariant Convolutional Networks for Pathloss Estimation, which leverages group equivariant convolutional networks to improve pathloss estimation accuracy. Categorical Equivariant Deep Learning is also noteworthy as it unifies various equivariant networks and proves the equivariant universal approximation theorem.